Perception, Representation, Language
Perception, Representation, Language
Perception, Representation, Language
representation,
language
frank jackson
http://www.amherstlecture.org/
Preferred citation
Jackson, Frank. Perception, Representation, Language. The Amherst Lecture in Philosophy 7
(2012): 117. <http://www.amherstlecture.org/jackson2012/>.
Abstract
We use sentences to make claims about how things are. Many of these claims come, in one
way or another, from how our perceptual experiences represent things to be. I argue in the
lecture that this tells us that very often we need sets of centered worlds (instead of sets of
worlds) to capture the contents of our sentences. I review some of the implications of acknowledging this for the debate over proper names and the referential behavior of the word
water.
The Amherst Lecture in Philosophy (ISSN: 1559-7199) is a free on-line journal, published by the
Department of Philosophy, Amherst College, Amherst, MA 01002. Phone: (413) 5425805.
E-mail: alp@amherst.edu. Website: http://www.amherstlecture.org/.
Copyright Frank Jackson. This article may be copied without the copyright owners permission only if the copy is used for
educational, not-for-profit purposes. For all other purposes, the copyright owners permission is required. In all cases, both
the author and The Amherst Lecture in Philosophy must be acknowledged in the copy.
We language users have views about how things are and, on occasion, want to share them
with our fellow language users. Thats why we produce sentences like Global warming is
a myth, Watch out, that plate is hot, and Arsenic is poisonous. Thats why we write
histories and publish physics textbooks and why I am giving this lecture. That is to say,
a principal use for words and sentences is to give information, to make claims about how
things are, to represent the world and the things in it as being a certain way. Here we have
three different phrasings designed to get across the same basic idea you can say it in terms
of information, or in terms of representation, or in terms of making claims.
How could a word or sentence a certain physical configuration on a page, or on a
computer screen, or encoded in a sound wave give information about the way things are? In
the broad, we know how the answer has to go. There are functions that go from the physical
structures in question to ways things might be, functions we grasp as competent users of the
language in question, and we exploit our knowledge of the functions to extract the information, to find out how things are being represented to be.1 Thus, there is a function that goes
from the sentence X is circular to the shape X has to have in order to be circular. Compe-
Of course, competence comes in degrees, but we will set this complication aside in what follows.
tent English speakers grasp that function and this is how, when they hear X is a circular uttered assertorically, they know what the speaker is saying about Xs shape it is the right way
for circular to apply to X. And they know what that is it is to be circular.2 Again, there is
a function that goes from Global warming is a myth to a way the world might be, namely,
a world where global warming is a myth. English speakers grasp that function and this is why,
when they wish to claim that global warming is a myth, they produce Global warming is a
myth to one or another audience.
We can say all this in terms of reference, but here I think we have to be a little careful.
The term reference gets used in (at least) two different though related ways in discussions
in the philosophy of language, and sometimes they get mixed up and confusion results.
Suppose I ask you Do you know the reference of the phrase The top of the stock market? You might say, Of course not. Thats a very valuable piece of information. It tells one
when to buy and sell. If I knew the reference of the top of the stock market, I would be holidaying in some luxury resort paid out of the profits of my trades, not having a conversation
with you here in Amherst. But of course there is a sense in which you do know the reference
of the top of the stock market. You know what it takes to be the top of the stock market.
You know how something has to be to be the top of the stock market. What it takes is for it
to be the peak, to put it in graphical terms, and you know that without knowing where the
peak comes. Indeed, your knowledge of the reference of the the top of the stock market in
this second sense explains how you know that knowing its reference in the first sense would
be such a valuable piece of information.
Our concern in this lecture will be with the second sense of reference: the what it
takes sense, the how something would have to be sense. This is the sense that feeds most
directly into any account of how words and sentences give putative information, and of what
a speaker is claiming about how things are when they use a word or a sentence. For this is the
sense which connects the words and sentences that come from our mouths, pens and computer keyboards, with how things are. This is the sense which connects with our earlier talk
Philosophers disagree about the conditions under which someone knows what it is to be circular Does
one need to be able to give the geometers definition? Must one be able to recognize a circle?... but that,
I insist, despite dissent in some quarters, is no reason for doubting that pretty much anyone reading this
lecture knows what it is to be circular.
of functions. The functions that underpin the way we extract information, know whats being claimed, know how things are being represented to be, are reference relations understood
in the what it takes sense of reference.
I hope the foregoing strikes you as common sense about words and sentences, dressed up
in a bit of jargon, or, if you like, as folk theory spelt out a bit. Do any of us need to do a course
at university to know that sentences give information, and to know what information a sentence like I am hungry imparts to English speakers? However, despite its commonsensical
flavor, it raises some important and tricky questions. Heres an example.
I am of course referring to Hilary Putnams famous Twin Earth thought experiment in The Meaning of
Meaning, in Mind, Language and Reality (Cambridge: Cambridge University Press, 1975), 215271.
water in the way that means that XYZ isnt the right way for water to refer to it. What
that way is exactly, we discuss later; for now, what is important is that there is such a way.
Others will object to a presumption in the penultimate paragraph. I presumed that, as a
rule, people in the early 1700s were justified in much of what they claimed about how things
are when they used the word water, and that any story about the reference of the word
water (understood in the sense of reference tied to information) should respect this fact.
Maybe what some in the early 1700s claimed about how things are when they used the word
witch wasnt justified, but surely the situation was very different for the word water. This
presumption underlay the earlier claim that we should not read their assertions as ones about
H2O. For in the early 1700s, people werent justified in believing, for example, that plants
need H2O to survive, or that there is H2O over the horizon. Some may object (have objected
in fact) that in making this presumption I confuse epistemology with philosophy of language.
I agree that epistemology and philosophy of language are different subjects in a departments list of offerings, but that doesnt mean that what we say in the philosophy of language
class isnt answerable to epistemic constraints. We use words and sentences to express how
we take things to be, as we said right at the beginning. I, along with thousands, am convinced
that Plato was a great philosopher, and that my conviction is fully justified: it is certainly true
and something I know about the world I inhabit. I use words and sentences to express this
conviction, most obviously the words Plato was a great philosopher (the words that in fact
occur in the preceding sentence). This means that what I claim about how things are using
the sentence a question about the referential properties of my words has to be something
that is certainly true and known by me to be true. The same goes, I insist, for the use of the
word water in the early 1700s in sentences like Plants need water to survive. Any account of the reference of the word needs to be consistent with these sentences making claims
about how things are that were known to be true in the early 1700s.
Still others will object that people in the 1700s did know that water is H2O, and were
justified in believing that water is H2O.4 What in that case, I reply, were those experiments
in the late 1700s and early 1800s all about?
In phrasing matters this way, I follow the common practice of supposing that water is H2O. Maybe, strictly
speaking, we should say that water is composed of H2O, or is composed of hydrogen and oxygen in a ratio
So there really is a puzzle. I will tell you what I think the solution is about three quarters
of the way through this lecture.
Its content, for short, in much of what follows. To highlight the fact that there are other things one might
mean by content, I elsewhere call content in this sense ir-content. See Frank Jackson, Language, Names,
and Information (Chichester: Wiley-Blackwell, 2010).
than an SUV. And to classify or categorize is to make a partition. In the case of information
about our world, it will be a partition into the worlds which fall under the classification versus those that dont. In the case of kinds of cars, it will be a partition that divides the sedans
from, say, the SUVs and the pickups.
Of course thinking of information in terms of partitions among possible worlds poses
many questions. How do the possible worlds it trades in relate to the distinction between
whats conceptually possible versus whats metaphysically possible? Whats the metaphysical status of the non-actual possible worlds? How does the set of worlds thought of as the
informational or (representational) content of some given sentence relate to the sentences
meaning? These are challenging questions, but I want to insist, all the same, that there
is something basically correct about the set-theoretic/partition among possibilities way of
thinking about information and representation. Someone worried about whether a tumor is
malignant or benign wants to know which of two possibilities is actual.
The main message I want to get across in this lecture, however, is that, in very many
cases, we should replace the talk of partitions among possible worlds with talk of partitions
among centered possible worlds, and that doing this has significant repercussions for how we
get information from coming across sentences we understand. It also gives us a simple way
of resolving the puzzle we raised earlier about the word water, and in addition gives us an
attractive way of thinking about the informational role of proper names.
We will start with the need to traffic in centered worlds.
ing us of this, perhaps the best known players being John Perry and David Lewis.6 I accept
the conclusion of these arguments but I think the focus on them has been in some ways
unfortunate. It has lead some to think that we only need centered worlds for rather special
cases, whereas in fact the need for them is ubiquitous. Also, the arguments that convinced
us are often based on discussions of cases where someone knows everything there is to know
of the form The so-and-so is F but does not know which one of the various so-and-sos they
themselves are. And it is controversial what exactly we should say about these cases. The net
effect has been to make centered worlds and why we need them appear subjects of mystery
and controversy to some. Whereas, I want to say, we need them for something as everyday as
what happens when we have a perceptual experience, and that what we need that is, the
centered worlds arent at all mysterious. Lets start by looking at the first point.
To have a perceptual experience is to be in a state that represents things as being a certain way: that the phone is ringing in the next room, that the train I am sitting in has just
started to move, that the computer I am looking at has gone to sleep, that there is a spider
crawling up my leg. In that sense, a perceptual experience carries putative information, and
does so as part of its very nature qua experience. We need to learn how a sentence in some
given language represents things to be; we dont need to learn how our senses represent
things to be.7 This means that, for any experience, we can distinguish the ways things might
be that are in accord with the experience from those that arent. Experiences partition ways
things might be. But we cannot think of the partition in question as a partition among possible worlds. Thats because perceptual experiences represent how things are from a point of
view without representing how that point of view is other than that it is one where things
are thus and so vis--vis it. When I have an experience as of a cat in front of me, I am in a
state that represents that there is a cat in front of a certain point. My state, however, doesnt
represent where that point is. Equally, it doesnt represent when it is: 15:00 on 5 May 2012
See, e. g., John Perry, The Problem of the Essential Indexical, Nos 13, no. 1 (March 1979): 321;
David Lewis, Attitudes De Dicto and De Se, Philosophical Review 88, no. 4 (October 1979): 513543.
Though of course (i) we may need to learn what can be inferred from our experiences, and (ii) training
can help us have experiences with improved representational contents. Radiologists learn how to read
X-rays, and part of what happens is that their experiences change as they learn. They come to have, as we
might put it, representationally improved experiences.
or...? Nor does the experience represent who occupies the point. The only thing it says
about the point is that theres a cat in front of it. This means we cannot capture the content
of the experience with a set of worlds. Perhaps this is immediately obvious, but lets spell it
out by reviewing the three ways one might seek to capture the content with a set of worlds
and why each fails.
First, one might use a set of worlds, with the very same point in each having a cat in
front of it. But that would treat my experience as saying which point it is that has a cat in
front of it, and thats wrong. My experience doesnt say which point has a cat in front of it.
Second, one might use a set of worlds, each of which has exactly one point with a cat in front
of it, but with the point that has a cat in front of it varying from world to world. The idea
would be that by varying the point, one removes the objectionable feature of the previous
suggestion: the content no longer says which point has a cat in front of it. But that would
treat my experience as saying in part that there isnt any other point in space with a cat in
front of it, and thats wrong. My experience is silent on whether there is more than one point
with a cat in front of it. Finally, one might use a set of worlds, each of which has at least one
point with a cat in front of it, and with the point or points that have a cat in front of them
varying from world to world. But this would treat my experience as representing that there
is at least one cat somewhere or other, and my experience does more than this. It represents
that there is a cat in front of a particular point but without representing which point it is.
(There is a big difference between my saying that there is a cat in front of me and my saying
that there is a cat somewhere or other.)
How then do we get the content of the experience right? By thinking of the experience
as partitioning the points in a world into those that are the right way in that world to be as
the experience represents things to be versus those that arent. For each and every point in
each and every possible world, it will or wont be the case that the point has a cat in front of
it. Now take every pair of point and world such the point is the right way in the world, then
the resulting set is the content of my experience. And this set is, of course, a set of centered
worlds.
I have just told you why we need centered worlds to capture the content of one particular experience but I trust it will be obvious that the same argument could be made for any
experience. It is a general point about perceptual experiences that we need centered worlds
to capture how they represent things to be. Whats more, I hope I have de-mystified centered
worlds: they are simply point in a worldworld pairs. All the same, theres an issue that needs
attention. If experiences have centered content, what makes what could make, one might
ask an experience veridical? Thats the topic of the next section.
ence and the actual world. So, to say it in English (more or less), my experience as of a cat
in front of me is veridical just if, in the actual world, theres a cat in front of where I am located at the time I have the experience. (The fact that this sounds platitudinous is of course
confirmation.) Here I am fudging something. It is a nice question exactly how the creature
that has the experience locates the point from which its experiences represent things to be
(though this is not, as weve highlighted, part of the content). Is it the creatures body that
does the locating, or is it some part of the brains location thats crucial? Moreover, surely
we shouldnt be presuming materialism in this discussion. We want an account of how the
creature that has the experience locates the relevant point of view which makes good sense
in a Cartesian world.
I dont want to minimize the importance of these questions but I do think they are fairly
classified as questions for another time. Any plausible answer to them can be plugged into
the account of when an experience is veridical that I have just given.8
I am also ducking an important, related issue. One can have a veridical experience as of a single cats
being in front of one there really is one and only one cat in front of one without its being the case
that one sees that cat. Examples that go back at least to H.P. Grice, The Causal Theory of Perception,
Proceedings of the Aristotelian Society: Supplementary Volume 35 (1961): 121168, tell us this.
But it tells us something that is perhaps not so obvious. It tells us to expect that the contents
of very many sentences will be given by sets of centered worlds, and not by sets of worlds.
Indeed, we should expect sentences that have contents that can be given by a set of worlds
to be the exception.
I know this worries some people. It can appear to make sentences in general too selfregarding. When I produce a sentence with centered content, arent I, in one way or another,
saying how things are with me? I think that there is a sense in which this is true, but it is
not a sense that should worry us. It is exactly what we should expect. Take a sentence like
There is a cat in front of me. It is plausible that the content of that sentence is given by a
set of centered worlds, each center of which has a cat in front of it (at the time in question).
As we have seen, that kind of account is right for the corresponding experience, and it is
hard to believe that things would be different for the sentence that reports the experience.
But it does not mean that the content of There is a cat in front of me is a set of centered
worlds with a cat in front of the person producing the sentence. The sentence says nothing
about producers of sentences or the sentences they produce. It isnt about a sentence or the
producer of a sentence; it is about a cat. The role of the producer of the sentence is to locate
the center for those of us who come across the sentence. The producer of the sentence is a
center locator, as I will call it.9 If you want to know where the cat is, attend to where the
producer of the sentence is and work from there. Knowing this is part of knowing how me
works in sentences of this kind. But it isnt part of how the sentence represents things to be;
that is why the producer of the sentence isnt part of the content. Perhaps it is worth laboring this point by spelling out how information gets acquired when we come across sentences
with centered content. Here is how it goes for our example.
1. You come across a token of There is a cat in front of me.
2. You have reason to trust it (you have reason to think that Im not lying, my eyes
are working properly etc.).
3. Its content is a set of centered worlds with a cat in front of their centers.
And see Jackson, Language, Names, and Information. The role of the producer of the sentence in locating
the center for the sentence parallels the role of the persons having the experience in locating the center
for the experience.
10
The importance of doing this is highlighted by Robert Stalnaker, Our Knowledge of the Internal World
(Oxford: Oxford University Press, 2008), 50. I am not suggesting that he would agree with what I say.
Now we are able to solve the puzzle about water. How come the word water in the
early 1700s was used to give information about the watery kind and yet it fails to refer XYZ
on Twin Earth; how come XYZ isnt the right way for water to apply to it, despite its being
every bit as watery as H2O?
I think we use the word water for the watery kind we have often come across.11 The kind
isnt always liquid and all that; it isnt always, as we might put it, especially watery. But it is
the kind which is very often liquid and all that, and which we have encountered. If that is
how the word was used in the early 1700s, it is transparent how come it was a word for giving information about the watery kind. It is also transparent why the word (in our mouths)
doesnt refer to XYZ on Twin Earth. XYZ isnt the kind we Earthians encounter; its the kind
the Twin Earthians encounter. On this view, sentences containing water have centered
content. The word is a word for talking about a kind we have come across, a kind that stands
thus and so vis--vis ourselves. The content of, for example, Water is plentiful is, accordingly, the set of centered worlds with centers that have encountered a watery kind that is
plentiful.
The Interplay between the Set of Centered Worlds and the Center Location Rule
I have said that there are lots of sentences with centered content. I have given a few examples but I trust it will be obvious how to generate many more. As I said earlier, having centered content is the norm. I also said that, in order to get information from sentences with
centered content, you need to know how sentence tokens locate the centers in question. I
gave one example: the way the appearance of me in certain contexts tells one that the center is the producer of the sentence. An obvious topic for future work is giving accounts, for
a whole range of sentence constructions that have centered content, of how tokens of these
sentence constructions serve to locate the relevant centers. This lecture is not the place for
11
The we are those of us who use water in such a way that Putnam is right that XYZ on Twin Earth
doesnt fall within its extension. Of course a version of this idea is to be found in Putnam, The Meaning
of Meaning.
this exercise. However, it is, I think, worth noting how an account of the set of centered
worlds for a sentence can intersect with an account of the how the token sentence locates
the center.
Take the sentence It is hot 1km due North of me. One way to give its content is as
the set of centered worlds with centers 1km South of a hot region. The corresponding center
location rule will then be: the producer of the sentence token. However, one could equally
do it this way. The content is the set of centered worlds with hot centers, and the location
rule is: 1km due North of the producer of the sentence token. Surely, it is obvious that there
is no real difference between the two ways of spelling things out. I am sure that this is not an
isolated example. There will many cases where there are two different but equivalent ways
of skinning our cat. Shortly, I will offer a short account of the informational value of proper
names in terms of a set of centered worlds and a center location rule. I am sure that there are
different ways of specifying the set of centered worlds combined with correspondingly different location rules that deliver the same net outcome for the informational value of proper
names.12
Proper Names
People who utter sentences of the from N is F give information about how things are,
often information properly speaking, as is the case for those who utter Paris is pretty or
Plato was a great philosopher. You might think of the description theory of reference for
(proper) names and the direct reference theory of names as offering competing accounts
of this information (there are other ways of thinking of those two theories of course). The
description theory might be thought of as saying that the information is given by a set of
worlds where the so-and-so is F, and the challenge set by Saul Kripkes attack on the description theory is then to come up with a value for so-and-so that survives his objections.13
12
Having said this, recent conversations with Dan Marshall have impressed on me the advantages of adopting as an invariable center location rule: the producer of the sentence token, and making the appropriate
adjustments to the set of centered worlds. I have, however, not followed this policy in the section that
follows.
13
Saul Kripke, Naming and Necessity (Cambridge: Harvard University Press, 1972).
The direct reference theory might be thought of as saying that we should give up on trying
to save the description theory and instead affirm that the information is given by the set of
worlds, w, where N, that very thing and independently of what descriptions it satisfies in w,
is F. From the perspective of this lecture, both suggestions are highly dubious. The ubiquitous
role of experience in grounding our claims about how things are, including those we make
using names, strongly suggests that we should be looking for an account in terms of a partition in centered worlds, not in worlds be they specified la the description theory or la
the direct reference theory combined with an account of how names serve to locate the
relevant centers.
Here is one way to do this. Take the set of centered worlds for N is F to be the set of
centered worlds whose centers are F. Here we take a leaf out the direct reference theorists
book in the sense that there is no descriptive requirement being placed on the centers by
N. Then think of the token name as serving to locate those centers. How might it do
this? Think of how historians approach questions like Did Helen of Troy really exist? and
Are the Robin Hood stories about a real person, and if they are, was he the good guy the
stories make him out to be? Historians approach these questions by looking for informationbearing causal chains that run from tokens of the names in various texts back in time to persons named in such a way that they generated those chains. In doing this, they treat names
as building blocks in information highways. And this is indeed what we do every day with
names. This is how names get to be wonderful devices for passing on information about individual objects. That indeed is the underlying rationale for giving objects names in the first
place.14 Here we are taking a leaf out of the causal theory of reference for names. What we
have just being saying is not that different from the kinds of things one finds in Naming and
Necessity, and in the writings of some of those influenced by this book. (I am not saying that
this is how they think of matters.)
There is, however, one important difference. Many supporters of the causal theory of
reference for names insist that the causal facts that secure the reference of a proper name are
not common knowledge; thats typically how they distinguish themselves from supporters of
14
For more on this (but the basic idea can be found in many places, as you would expect given that it is folk
theory), see Jackson, Language, Names, and Information, ch. 5.
causal descriptivism. But this sits ill with the way the method is used by historians to answer
questions about Helen of Troy and Robin Hood. They dont feel the need to get advice from
philosophers of language. Or consider reviewers of their books. These reviewers express opinions about whether or not the authors have or have not shown that Robin Hood or Helen of
Troy or Job or ... really existed and did this, that or the other thing. But mostly the reviewers
are not experts in the philosophy of language; they are historians and writers, and moreover
they take it for granted (rightly) that both the books they are reviewing and their reviews of
them are accessible to reasonably intelligent readers quite independently of whether or not
they have done philosophy at university. And this is to take it for granted that reasonably
intelligent readers know what it takes to be the right kind of causal origin of the tokens of
Robin Hood and Helen of Troy, or, if it comes to that, Tony Blair. This is what I had in
mind when I said earlier, in a footnote, that I was offering a bit of folk theory.
Acknowledgments
I was delighted to be asked to give this lecture. In writing the lecture up for publication, I
have taken account of the helpful discussion at the lecture. I have also drawn on many discussions over many years, most recently after oral presentations of related material at Cambridge University, May 2011, and Adelaide University, March 2012.
References
Grice, H.P. The Causal Theory of Perception. Proceedings of the Aristotelian Society:
Supplementary Volume 35 (1961): 121168.
Jackson, Frank. Language, Names, and Information. Chichester: Wiley-Blackwell, 2010.
Kripke, Saul. Naming and Necessity. Cambridge: Harvard University Press, 1972.
Lewis, David. Attitudes De Dicto and De Se. Philosophical Review 88, no. 4 (October
1979): 513543.
Perry, John. The Problem of the Essential Indexical. Nos 13, no. 1 (March 1979): 321.
Putnam, Hilary. The Meaning of Meaning. In Mind, Language and Reality. Cambridge:
Cambridge University Press, 1975.
Stalnaker, Robert. Our Knowledge of the Internal World. Oxford: Oxford University Press,
2008.