Indexical thought
Frege cases:
Hesperus/Phosphorus etc.
A rational subject, S, may take different
(and possibly conflicting) attitudes
towards the judgment that a given
individual is F depending on how that
individual is presented.
 To account for that sort of situation,
Frege posited ‘modes of presentation’ in
addition to the reference of linguistic
expressions.

Frege’s constraint on ‘modes of
presentation’
If a rational subject can think of some object a
both that it is F and that it is not F, that means
that there are two distinct modes of
presentation under which the subject thinks of
a.
Modes of presentation are whatever is needed,
in addition to reference, to account for the
subject’s differential behaviour in such
situations.
Indexical modes of presentation
They are a special case.
First person vs third person modes of
presentation:
My pants are on fire
His pants are on fire
Modes of presentation: two
options
Semantic option : Modes of presentation are a
level of content additional to reference.
(Fregean ‘senses’ etc.)
Syntactic option : Modes of presentation are
not an additional level of content. They are
‘concepts’ qua vehicles of content. (Fodor,
Sainsbury & Tye, etc.)
Arguments for the syntactic
option

Millianism: Reference is all there is to
content. No ‘senses’.
 Mates’s cases : There being two different
words (eg ‘psychiatrist’/‘alienist’, or
‘Greek’/‘Hellene’) is sufficient to make
Frege cases possible.
 Even if there is a single word
(‘Paderewski’) in the language, Frege cases
will be possible if the subject associates
distinct ‘mental files’ with it.
Numerically distinct files
The two ‘Paderewski’ files contain distinct
bodies of information (about Paderewski the
musician and Paderewski the politician
respectively).
 But two distinct mental files may also
contain the same information.
 There being two distinct mental files is
sufficient to make it rational for the subject
to contemplate the possibility that one
object is F while the other (perhaps) isn’t.

Preliminary conclusion
Modes of presentation = mental files.
On the syntactic construal, mental files are
mental particulars.
They are not to be equated to the body of
information in the file.
The reference of mental files
Mental files serve to store information
about objects (so they are about objects
and ‘refer’).
 The reference of the file is not
determined by information in the file but
through relations to entities in the
environment in which the file fulfills its
function.

ER relations
The relations on which mental files are
based, and which determine their
reference, are epistemically rewarding
relations. They enable the subject to
gain information from the objects to
which he stands in these relations. The
function of the file is to store information
from these entities — information that is
made available through the relations in
question.
An example: demonstrative
files
The subject stands in a certain relation to
some object he is perceptually attending
to.
In virtue of that relation, the subject can
gain (perceptual) information about the
object.
The demonstrative file ‘that F’ serves as
repository for information gained in that
way.
Another example: the ‘self’ file
In virtue of being a certain individual, I
am in a position to gain information
concerning that individual in all sorts of
ways in which I can gain information
about no one else, e.g. through
proprioception and kinaesthesis.
 The mental file self serves as repository
for information gained in this way
(Perry).

Mental files and the
role/content distinction
Different types of file exploit different
types of contextual relation.
 The type of a file corresponds to its
function or role: exploiting a given ER
relation.
 This shows that there are two levels of
content after all: referential content (a
property of the file-token) and functional
role (a property of the file-type).

Two levels of content: the
Perry/Kaplan approach
If you and I both think ‘I am tired’, there is
a sense in which we think the same thing,
and another sense in which we think
different things.
 It would be misleading to say that the first
level (the level at which we think the same
thing) is ‘purely syntactic’ ; for what
characterizes that level is the function or
role of the files we deploy in our respective
thoughts. The function or role stays
constant : we both deploy a SELF file.

‘My pants are on fire’/’His
pants are on fire’

There is a sense in which we think the same thing,
and another sense in which we think different
things.
 Same truth-conditional content, but different
thoughts, with different behavioral effects.
 What characterizes the difference is, once again,
the function or role of the files that are deployed in
the respective thoughts: a SELF file or a
demonstrative file.
Summing up

Modes of presentation are mental files, construed
as mental particulars.
 They are not senses, but referential vehicles.
 Still, we must distinguish two levels of content, as
in Frege’s framework.
 The two levels we need correspond to Kaplan’s
character/content distinction: qua tokens mental
files refer, but qua type they possess a ‘character’
corresponding to their role or function.
The indexical model for
language
expression type
encodes
expression token
reference
contextual relation
The indexical model for
thought
mental file type
has the function of
storing information
derived through
mental file
token
contextual relation
reference
of file
Alternative approaches

Can we account for the cognitive
significance of indexical thoughts without
appealing to mental files and the
vehicle/content distinction ?
 Can we do it purely in terms of content ?
 Both the centered world framework and the
token-reflexive framework can be seen as
attempts to do that.
Alternative approaches (1)
Centered contents
Instead of introducing the vehicles into the
picture and endowing them with functional
significance, Lewis proposes to make the
contents themselves more fine-grained by
characterizing them as sets of centered
possible worlds rather than as sets of
possible worlds tout court. Centering the
possible worlds on an individual at a time
gives us the subjective perspective which is
the hallmark of indexical thought.
Centered contents
Centered contents are not classical propositions
(which only require a possible world to determine
a truth-value), but relativized propositions.
They determine a truth-value only when evaluated
with respect to an appropriate index, containing
the thinking subject and the time of thought in
addition to the world in which the thought occurs.
The content is a property of thinker-time pairs, not a
classical proposition; and it is evaluated ‘at’ the
individual/time of the index.
‘That object is round’

The content is the set of all possible worlds
that are centered on an individual who is
seeing a round object. That is the property
of seeing an object that is round.
 The subject who judges ‘that object is
round’ self-ascribes that property (‘force’
component).
 The self-ascription is true iff the subject of
thought is actually seeing a round object at
the time of thinking.
Lewisian Descriptivism

The objects that are represented in the
content of the thought (e.g. the round
object) are represented descriptively.
 They are described as bearing such and
such relations to the ‘center’, i.e. to the
subject of thought (at the time of thinking).
 The acquaintance relations are ‘internalized’
and reflected into the content of one’s
thought.
Centered contents and
Descriptivism

If I see something, I think of it descriptively
as ‘what I see’ – the object that bears a
certain perceptual relation to me. (Searle,
Jackson...)
 In general, the objects we are acquainted
with are represented descriptively as
bearing such and such acquaintance
relations to the subject.
The limits of Lewisian
Descriptivism
A qualification: The description ‘what I see’ is not
fully appropriate.
 In ‘what I see’, there is an occurrence of the first
person. It corresponds to the subject in the
contextual index, and the subject is not
represented in the content : it is externalized and
directly provided by the context.
 The subject is not represented but, qua subject of
the thought episode, it is involved pragmatically in
the process of ‘self-ascription’ through which
Lewis characterizes the attitudinal mode.

The Lewisian asymmetry

Those objects of thought that belong to the
contextual index are treated completely differently
than the objects of thought that are represented in
the content of the thought.
 The objects of thought in the content are
represented descriptively as bearing such and such
relations to the ‘center’.
 In contrast, the entities in the contextual index are
‘externalized’ and directly provided by the
context.
Problems with the view
• I deplore the descriptivist construal of the
content of thought and the internalization of
acquaintance relations.
• I also deplore the asymmetry and its
solipsistic/idealistic flavour.
(As Chisholm puts it, ‘There is one sense in
which the believer can be said to be the
object of his believing’.)
Multiple anchors

In the Lewis-Chisholm framework, everything is
thought of descriptively, except for a single
element which is externalized and serves as
universal anchor for all the content.
 Why not appeal instead to multiple anchors,
corresponding to all the acquaintance relations in
which we stand to objects of thought?
 Multiple anchors are precisely what the mental file
framework gives us, thus doing away with both
the asymmetry and Descriptivism.
Multiple anchors in the centred
world framework

Multiple anchoring can be achieved also in
the centred world framework by putting
sequences of objects in the contextual index
(Ninan’s ‘sequenced worlds’)
 If we do that, however, we have to
introduce modes of presentation (mental
files) into the picture. We lose the benefit of
Lewis’s successful appeal to the attitudinal
mode to capture the first person perspective.
Sequenced worlds: the mode of
presentation problem

If the objects of thought are fed into the
contextual index, what will determine how
the objects in question are thought of ?
 The descriptivist packs the modes of
presentation into the content, but if we don’t
do that, we need some other way of pairing
the objects with the right modes of
presentation.
The attitudinal mode and its role
in the Lewis framework
In Lewis’s original framework, there is a
(nondescriptivist) way of pairing the subject in the
contextual index with the right mode of
presentation (the ‘self’ mode of presentation).
 An attitudinal state is analysed into content and
mode. The content, for Lewis, is a property. The
belief mode itself is analysed by saying that to
believe a content (analysed as a property) is for
the subject of thought to ‘self-ascribe’ that
property.

What is it to ‘self-ascribe’ a
property?


It is not just to ascribe that property to oneself.
There are different ways in which one can ascribe
a given property to oneself, corresponding to
different modes of presentation of oneself.
 The thinker can think of himself that he is tired,
when seeing himself, looking tired, in the mirror
(without realizing that he is looking at himself).
Or he can think that he is tired on the grounds that
he feels tired. Only in the latter case does Lewis
analyse the content of the thought as the property
of being tired, which the subject ‘self-ascribes’.
What is it to ‘self-ascribe’ a
property?
There is no possibility of ‘self-ascribing’ a
property under a 3rd person mode of presentation
of oneself (say, as the man seen in the mirror).
 It is the attitudinal act of ‘self-ascription’ itself
which determines a particular mode of
presentation of the subject to whom a property is
ascribed.
 In other words, the first personal mode of
presentation is built into the self-ascriptive
relation.

What happens when we enrich
the contextual index by feeding
it a sequence of objects?

Each object in the sequence can be thought of
under a number of distinct modes of presentation.
So we need a way of pairing the objects with the
appropriate modes of presentation.
 But we can’t use the Lewis trick. There is a single
self-ascriptive mode. On that mode we can base a
special mode of presentation in Lewis’s original
framework because a single individual occupies
the center, and it is that individual that we need to
pair with the right mode of presentation to avoid
mirror-type counterexamples.
What happens when we enrich
the contextual index by feeding
it a sequence of objects?

When we multiply the individuals in the
contextual index, what we need is not a single
mode of presentation, but a sequence of modes of
presentation corresponding to the sequence of
objects. Appealing to the attitudinal mode is of no
use here!
 Conclusion: once we revise the framework so as to
get rid of the asymmetry, we can no longer
account for cognitive significance purely in terms
of content. We need to add modes of presentation.
Alternative approaches (2)
Reflexive contents
The token-reflexive framework (Searle, Perry,
Garcia-Carpintero, Higginbotham) also
appeals to a special sort of truth-conditional
content, in order to deal with indexical
thought.
Alternative approaches (2)
Reflexive contents

Objects are thought of as bearing certain relations
not to the subject at the time of thinking but to the
occurrence of the thought in which they are
represented. Each thought or utterance is therefore
ascribed a reflexive content that is about that
thought or utterance itself.
 For example, an occurrence u of ‘I am tired’ in
speech or thought means something like ‘the
utterer/thinker of u is tired at the time of u in the
world of u’.
A problem for Reflexivism
If I say or think ‘I am tired’, and this is analysed
as ‘the utterer/thinker of u is tired at the time of u
in the world of u’, then I have referred to myself
under the descriptive-relational mode of
presentation ‘the utterer/thinker of u’.
 Every object of thought is referred to under such a
descriptive-relational mode of presentation which
exploits the object’s relation to u. But what about
u itself ? Under which mode of presentation is it
referred to ?

‘This very occurrence’
One option for the reflexivist is to say that u is
thought of as ‘this occurrence’, where ‘this’ is
understood reflexively (‘this very…’)
But such a reflexive mode of presentation cannot
itself be given a descriptive-relational analysis.
If ‘this occurrence’ were analysed as ‘the occurrence
that is identical to this’, we would be using the
analysandum, namely the reflexive ‘this’, in the
analysans. If it were analysed as ‘the occurrence
that is identical to u’, we would be back to where
we started and would still be in need of a mode of
presentation for u.
An alternative: Super-direct
reference
Reminiscent of Russell’s strong notion of
acquaintance (with ourselves and our sense
data).
 In super-direct reference, there is no mode
of presentation. The object itself is directly
recruited as a thought constituent.
 This of course cannot be done with many
objects, but with mental occurrences
arguably it can.

Super-direct reference

Super-direct reference is supposed to be
‘transparent’ (in contrast to ordinary direct
reference): under super-direct reference, no Frege
cases are possible.
 The idea of super-direct reference tends to surface
in discussions of phenomenal concepts. See e.g.
Chalmers’s statement that, in the phenomenal
case, ‘the referent of the concept is somehow
present inside the concept’s sense in a way that is
much stronger than in the usual case of direct
reference’ (Chalmers 2003).
Reflexivism in Lewisian clothes
This idea can be couched in Lewis’s framework, by
externalizing the occurrence u and letting it be
directly provided by the context. Everything is
then described relative to u, but u itself is given, it
is not represented.
In this framework as in Lewis’s:
(i) acquaintance relations are internalized : relational
descriptions provide the modes of presentation
under which objects are thought of.
(ii) there is an exception : the occurrence in terms of
which everything (else) is descriptively
characterized!
Reflexivism in Lewisian clothes
On this mixture of the two frameworks
(centered worlds and reflexivism):
• The content of a mental occurrence is a
property of occurrences.
• That content is evaluated with respect to a
contextual index containing the occurrence
itself.
Reflexivism in Lewisian clothes

To judge something by assertively tokening a
certain representation is to ascribe to the token the
property that is its content.
 Here reflexivity is guaranteed by the pragmatic
architecture of the act of judgment.
 When you think ‘I am tired’, the content of the
thought is the property an occurrence has just in
case the thinker of that occurrence is tired at the
time of the occurrence in the world of the
occurrence. To think the thought (or to think it
assertively) is to ascribe that property to the
current occurrence u you are producing.
Problems with the view (the
same as for the Lewis picture)

Descriptivism: Everything is thought of
descriptively, except for a single element which is
externalized and serves as universal anchor for all
the content.
 Dramatic asymmetry between different objects of
thought, i.e. between the universal anchor and the
rest, motivated by some kind of extreme Cartesian
picture. (The mind’s transparent access to itself
serves as the foundation for all our knowledge.)
My picture
• Objects are thought of (either descriptively
or) under modes of presentation which are
mental files.
All objects -- so no asymmetry!
• Mental files are based on acquaintance
relations, but to think of an object through a
mental file you don’t have to think of the
relation on which the file is based. No
descriptivism!
Attunement (Perry)
Objection: Acquaintance relations are not
external to the mind. We we are ‘attuned’ to
the relations which determine what we’re
thinking about. Yes but:
‘there is a difference between being able to
think of a thing or person in virtue of some
role it plays in one’s life, and being able to
articulate that role in thought or speech and
think of it as the thing or person playing
that role in one’s life’ (Perry 1997)
Reflexivism without
Descriptivism
To protect Reflexivism from the charge
of Descriptivism, one can introduce (as
Perry actually does) a multi-level
framework, with the reflexive content
occurring at one level and the referential
content at another.
 Then one can say that the reflexive
content is not represented even though
the ‘attunement’ relation holds.

Getting rid of Descriptivism
The recipe:
• go two-dimensional
• distinguish between two distinct ‘grasping’
relations.
Attunement counts as the grasping relation
appropriate to reflexive content (vs
referential content)
Attunement again
‘Attunement to the relation that our self-notions have
to ourselves, or our perceptions have to the object
they are of, does not require belief or thought
about the relation ; it requires know-how, not
knowledge that’ (Perry 2012 : 99).
It’s a matter of function or role.
- Function of role of what?
- How can we answer that question without bringing
the vehicles into the picture? (Indeed, Perry
himself appeals to mental files: they are ultimately the tokens in his token-reflexive framework.)
Conclusion
No clear alternative to the mental file
framework.