1. The Concept of Analysis
1.1 The Classical Concept of Analysis
( (1) The relation of analysis is transitive and asymmetric:
(a) If concept A can be analyzed in terms of some set of concepts, B,
and if all of the concepts in set B can be analyzed in terms of the
concepts in some set of concepts, C, then concept A can be analyzed in
terms of the set C of concepts.
(b) If concept A can be analyzed in terms of some set of
concepts, B, then none of the concepts in set B have an analysis that
involves concept A.
(Another way of saying that the relation of analysis is asymmetric is
to say that the relation of analysis has an intrinsic direction.)
An immediate consequence of (1) is
(2) In analyses, there can be no looping around in circles.
Another feature of the classical concept of analysis
is
(3) No infinite regress involving concepts that are analytically more
and more basic is possible.
A consequence of (2) and (3) is
(4) If there are any concepts, then there are concepts that are
analytically basic.
1.2 Alternatives to the Classical Concept?
An alternative view is that classical analysis is impossible. What one has is just a web of interrelated concepts, none of which is prior to any other. So analysis is merely relative. Any concept can be analyzed; it is just a matter of choosing the right conceptual framework. (Compare Nelson Goodman, discussed below.)
1.3 Goodman’s New Riddle of Induction: The Problem of Gruesome Concepts
1.3.1 Goodman’s Own Formulation2. Criteria for Analytically Basic
Concepts and Terms
2.1 Analytically Basic Concepts and Terms and the Verbal Indefinability Criterion
It is natural to divide terms
and concepts into logical terms and concepts and non-logical, or
descriptive, terms and concepts. It seems to me that very
different accounts have to be offered of the explanation of logical
concepts and terms on the one had, and descriptive concepts and terms
on the other. In what follows I shall consider only the case of
non-logical terms and concepts. For the sake of brevity, I shall
generally simply say “terms or concepts”, but the discussion that
follows should be taken throughout as concerned only with
non—logical/descriptive terms and concepts.
Where should analysis of
descriptive terms or concepts start? Of the descriptive terms or
concepts that one has, which of them should form the starting point for
the analysis of other terms or concepts? Which descriptive terms
or concepts are analytically/semantically basic?
The meaning of many descriptive
terms can be explained in terms of other terms. But in the case
of some descriptive terms, verbal definition does not seem
helpful. One needs, it certainly seems, to acquire an
understanding of some descriptive terms by what is often referred to as
"ostensive definition" – by being exposed to situations where the term
is actually being used.
But what counts as an "ostensive
definition"? One view is that one has an ostensive definition of
a term where a person is perceptually acquainted with the property or
relation in virtue of which the term in question applies to a
situation. As a characterization of ostensive definition, this is
probably fine. However, so understood, many terms that can be
ostensively defined can also be defined verbally. If there is to
be a class of analytically/semantically basic terms in a strong sense,
it will have to consist of terms that can only be defined
ostensively. Is there such a class?
The idea behind this question can
be brought out if one compares, say, the terms "car", or "table" with
the term "red". Most people will acquire an understanding of the
terms “car” and “table” ostensively, but that understanding can also be
acquired via a purely verbal definition. In this respect, such
terms seem to differ from a term like "red", since the meaning of the
term "red", it might seem, cannot be conveyed via a purely verbal
definition.
This suggests the following thesis:
The Verbal Indefinability y Criterion
A term is semantically basic if and only if the meaning of the term cannot be conveyed to another by a purely verbal definition.
2.2 The Immediate Perceivability Criterion
The above discussion suggests,
however, another criterion. This other criterion involves a
distinction between (a) direct, or immediate, or non-inferential
perception, on the one hand, and (b) indirect, or mediate, or
inferential perception, on the other. Exactly how this
distinction is best drawn is a somewhat tricky matter, as we shall see
in a later seminar, but here we can think in terms of a distinction
between (a) perceptual beliefs that involve no inference from other
perceptual beliefs and (b) perceptual beliefs that do involve inference
from other perceptual beliefs.
Consider a case where you are looking at an orange. One
perceptual belief that you acquire in such a situation is the following:
(a) The belief that there is a certain type of three-dimensional object
in front of you – namely, an orange.
Does that perceptual belief involve inference or not? Initially
many people would be tempted to say that it does not involve any
inference. But in thinking about that question, it is important
to note that the term “inference” here is used to indicate a certain
sort of causal relationship between beliefs, and does not imply the
presence of any thoughts, or any conscious process of inference.
With that in mind, consider the following, closely
related belief:
(b) The belief that there is a round, physical expanse that is orange
in color in front of one.
When one acquires the belief that there is a certain type of
three-dimensional object in front of you – namely, an orange – it will
typically be the case both that
(1) One also acquires the belief that there is a round, physical
expanse that is orange in color in front of one
and that
(2) One would not have acquired the first of those beliefs if one had
not acquired the second.
The upshot is that while the belief that there is a
round, physical expanse that is orange in color in front of one may be
a non-inferential perceptual belief, the belief that there is a certain
type of three-dimensional object in front of you that is an orange
cannot, it would seem, be a non-inferential perceptual belief.
This suggests the following thesis:
The Immediate Perceivability Criterion
A concept is semantically basic if and only if the concept enters into a proposition that can be the content of an immediate perceptual belief.
But, as we shall now see, this criterion appears to be unsound.
2.3 What Kinds of Terms Are Semantically Basic?
Consider the term "red". Does that term really present us with a case where verbal definition is impossible? There are two closely related lines of thought that suggest that terms such as "red", when applied to physical objects, can, in appropriate circumstances, be verbally defined:(1) The Blind, or Color-Blind Person;
(2) A World without Red Objects.
2.3.1 The Blind, or Color-Blind Person
Consider the case of the
person who is red/green color-blind. Couldn't one teach such a
person, by appropriately stimulating his or her brain, what it was like
to have an experience of the red variety, or an experience of the green
variety? And then, having done that, couldn't one explain what it
was for a physical object to be red, and what it was for it to be
green? For in the case of a color-blind person, one could say
that redness, understood as a property of physical objects, is a
property of the surface of objects – just as blueness is – the
difference being that redness gives rise, under normal conditions, to a
different sort of experience – namely, an experience of the red variety.
In the case of a person who,
rather than being color-blind, was blind from birth, one couldn't
compare redness with other color properties that the person was
familiar with. But even with that possibility removed, couldn't
one still explain what it is for a physical object to be red simply by
reference to the sort of experience – an experience of the red variety
– that a physical object gives rise to under normal conditions, and
where one gives the blind person an understanding of what such
experiences are like by appropriately stimulating his or her brain?
In both cases, one would have to
explain what normal conditions were. For without the reference to
normal conditions, the resulting account would not be satisfactory: a
white object illuminated with red light will give rise to an experience
of the red variety, but what one wants to say in such a case is merely
that the white object looks red under those circumstances, and not that
it is red.
How can the concept of normal
conditions be explained? Exactly what the right answer is, is a
slightly tricky question, but one can certainly see some promising
starting points. Thus, one possibility would be to equate normal
conditions in the case of color perception with bright sunlight.
Another idea is to equate normal conditions with those under which one
can make maximal color discriminations. (It is possible – though
not, I think, likely in the case of color – that precisely what
conditions make for maximal discrimination depends upon the properties
that one is perceiving.)
2.3.2 A World without Red Objects.
Consider a world that contains
no red objects. Once again, one could appeal to the possibility
of directly stimulating a person's brain in such a way as to produce
experiences of the red variety. But there is another possibility
that would be less invasive. Stare at a green object, and then
look at a white background. The result is a red
after-image. One could then define the term "red", as applied to
physical objects, as picking out that property of objects that, under
normal conditions of observation, gives rise to an experience that has
the quality that characterizes red after-images.
The conclusion, in short, seems
to be, first, that a person who never saw red physical objects could
nevertheless come to understand perfectly well what it is for a
physical object to be red, and secondly, that even in a world devoid of
red physical objects, it would be possible to come to understand what
it would be like for there to be red objects. Understanding what
it is for a physical object to be red does not presuppose, in short,
either acquaintance with that property, or even that there is any
instantiation of that property, anywhere, at any time.
Consider, in contrast, redness as
a property of sense experiences. If one had never had that type
of experience, it would seem that no verbal definition could supply one
with an understanding of what the experience would be like.
2.3.3 The Possibility of Acquiring an Understanding of Semantically Basic Terms without Having any Relevant Experiences
It is very natural at this point to
conclude that one cannot understand what it is to have an experience of
the red variety unless one has had such an experience, and that, in
general, one cannot understand the meaning of any semantically basic
term – or grasp any semantically basic concept – without having an
experience of the relevant property or relation. That claim,
however, is exposed to a certain objection – namely, that a person
might come to acquire the concept of an experience of the red variety
if his or her brain were changed so that it was similar in the relevant
respect to the brain of someone who did have the concept of an
experience of the red variety. So isn’t it the case that one
could in principle understand semantically basic terms, and grasp
semantically basic concepts, even if one had never had any experience
of the relevant property or relation?
In a certain sense, such a person
would have the relevant concepts, and would understand the term “red” –
both as a term applied to physical objects, and as a term applied to
experiences – since, if confronted with a red physical object, or if he
or she had the relevant sort of experience, that person would be able
to say right off that an apple was red, or that he or she was having an
experience of the red variety. But I am inclined to say that
until such a person either (1) has the relevant experience, or (2)
imagines having such an experience, he or she does not understand what
redness is, though he or she does possess the relevant concept.
2.4 Direct Acquaintance and Semantically Basic Terms and Concepts
Recall the following criterion:
B. The Immediate Perceivability Criterion
A concept is semantically basic if and only if the concept enters into a proposition that can be the content of an immediate perceptual belief. In his book Perception, Frank
Jackson defended the view that the immediate objects of perception are
sense data. If Jackson were right about that, then the Immediate
Perceivability Criterion would, I think, be correct. But while I
think that Jackson is right that the immediate objects of awareness are
mental objects, I am inclined to think that it is wrong to apply the
term “perception” to awareness of one’s own mental states, or parts
thereof. My reason for disagreeing with Jackson on this matter is
that I am inclined to accept some form of the causal theory of
perception, as defended by H. P. Grice, according to which one does not
perceive an object unless that object causally gives rise to an
experience.
Because of this, I think that the
immediate perceivability criterion needs to be replaced by one that is
formulated either in terms of direct awareness, or in terms of direct
acquaintance.
Formulated in terms of direct
awareness, the criterion can be stated as follows:
C. The Direct Awareness
Criterion
A concept is semantically basic if and only if the concept applies to
something by virtue of that thing’s having some property such that one
can be directly aware of instances of that property (and similarly for
relations).
What is direct awareness? The answer is roughly as follows:
(1) S is aware of x = def.
(a) S has the thought or the belief that x exists.
(b) That thought or belief is caused by x.
(2) S is indirectly aware of x
= def.
(a) S is aware of x
(b) There is some y such that S is aware of y, and where x caused S to
be aware of x by causing S to be aware of y.
(3) S is directly aware of x =
def.
(a) S is aware of x
(b) There is some causal process running from x to S’s awareness of x,
and that does not involve any thought or belief that y exists, where y
is anything that is distinct from x
Note that these definitions allow
it to be the case that there is something that, at a given time, one is
aware of both directly and indirectly.
Alternatively, formulated in
terms of direct acquaintance, the criterion can be stated as follows:
D. The Direct Acquaintance
Criterion
A concept is semantically basic if and only if the concept applies to
something by virtue of that thing’s having some property such that one
can be directly acquainted with instances of that property (and
similarly for relations).
But what is direct
acquaintance? The notion of direct acquaintance presupposes
the idea of phenomenal, sensuous, non-dispositional, intrinsic,
qualitative properties, including color properties. In the case
of color properties, it is natural to think of them as located on the
surfaces of objects, since that is where they appear to be.
Physics, however, provides very good reason for believing that
phenomenal, sensuous, non-dispositional, intrinsic, qualitative color
properties are not in fact located on the surfaces of physical
objects. If they exist, then, it seems that they must instead be
properties that are located within experiences.
If that is where phenomenal,
sensuous, non-dispositional, intrinsic, qualitative properties are
located, then they are spoken of as qualia. That there are qualia
is a controversial question in contemporary philosophy of mind.
If qualia do not exist, then there may be nothing to be directly
acquainted with, though some philosophers would say that one can be
directly acquainted not just with property instances, but also with
properties. By contrast, the notion of direct awareness is such
that the existence of phenomenal, sensuous, non-dispositional,
intrinsic, qualitative properties is not required in order for there to
be something of which one is directly aware.
In any case, given the idea of
phenomenal, sensuous, non-dispositional, intrinsic, qualitative
properties – ‘qualitative properties’, for short – one can give the
following definition of direct acquaintance:
S is directly acquainted with x = def.
(a) S has an indexical thought or belief about x,
(b) That indexical thought or belief is caused by x,
and
(c) x is a phenomenal, sensuous, non-dispositional, intrinsic,
qualitative property, either simple or complex.
Why “either simple or
complex”? The reason is that one’s visual experiences, for
example, typically involve instances of color properties that stand in
spatial relations to one another. So among the properties,
instances of which one can be directly acquainted with, is, for
example, the property of consisting of an instance of qualitative
greenness between two instances of qualitative redness
2.5 Strong Ostension Versus Weak Ostension
2.5.1 Ostensive Definition Based upon Erroneous Perception
In addition to
the arguments set out in sections 2.3.1 and 2.3.2, there is a third
consideration that throws light upon ostensive definition, and which is
as follows.
Regardless of whether
there are any red objects in the world, there may be objects that look
red under various circumstances, and if the term "red", as a predicate
of physical objects, were applied to such objects, and if there were,
in addition, a mistaken belief that the conditions were normal, then a
person would presumably acquire the concept of what it is for an object
to be red, even though he or she had never been exposed, not to any
object that really was red, but only to objects that looked red.
This could never happen, in contrast, in the case of "red" where one is
dealing with a predicate of sensory experiences. So there is a
distinction between terms that can be ostensively defined in
circumstances where the term does not truly apply to the object in
question, and terms that can only be ostensively defined in
circumstances where they truly apply to the objects in question.
It may be a good idea,
accordingly, to introduce the following distinction between weak
ostensive definition and strong ostensive definition:
A term, T, is defined via weak
ostension if and only if a person acquires an understanding of the
meaning of the term by seeing it used in situations to which it does
not truly apply. (The situations may very well be ones where its
use is reasonable, or appropriate, given what the user knows or
justifiably believes.)
(2) A term, T, is defined via strong
ostension if and only if a person acquires an understanding of the
meaning of the term by seeing it used in situations to which it does
truly apply.
2.5.2 Some Possible Theses Concerning Definition Via Ostension
Given this distinction, consider whether any of the following theses are true:
1. If a term is definable only via ostension, it is definable only by strong ostension, and not via weak ostension.
2. If a term is definable by strong ostension, but not via weak ostension, then it is definable only via ostension.
3. A term is definable only via ostension if and only if it is definable by strong ostension, but not via weak ostension.
4. Some terms can be defined only via strong ostension.
Since thesis 3 is the conjunction of theses 1 and 2, thesis 3 can only be true if both 1 and 2 are true. I shall argue that 1 is true, but that 2, and so 3, are false. I shall also argue that thesis 4 is true.
An Argument for Thesis 1
To think about whether this
first thesis is true, consider a term, T, which can be defined via weak
ostension, so that a person can come to understand the term T by seeing
it used in certain situations, even though the term T does not truly
apply to anything in that situation. If this is the case, then,
in the first place, there must surely be something that is true about
the situations in question that makes the use of the term T appropriate, or reasonable. If so, then it
would seem that one should be able to describe what it is about the
situations in question that makes it appropriate to use the term T,
even though it does not actually apply to anything in those
situations. So let us use the term S to refer to whatever that
relevant feature is.
Consider a concrete case.
Earlier, we saw that one can acquire an understanding of the term “red”
as applied to physical objects in situations in which no red physical
object was present. But what was present in all such cases was an
experience of the red variety. So what S would be in this case
would be a term – call it “red*” – that picks out qualitative redness,
where the latter, it turns out, is a property of parts of one’s visual
field.
But if the presence of
qualitative redness is to make it appropriate,
or reasonable, to believe that
there is a red physical object that one is perceiving, must there not
be some relation between the term “red*” that picks out the relevant
feature and the term “red”, in virtue of which the presence of
qualitative redness – the property picked out by the predicate “red*” –
raises the likelihood (at least
to a sufficient degree) that some red physical object is present?
But how can this be the case unless there is an analytical connection between the
property of being red* and the property of being red? The term
“red*”, however, is analytically basic, and cannot be analyzed, and so
there can be an analytical connection between the terms “red*” and
“red” only if the term “red” is analyzable, and analyzable in a way
that involves the term “red*”. So the term “red’ must be verbally definable.
It seems to me that this argument
is a perfectly general one, so that whenever a term T is definable via
weak ostension there must be a term S that, first of all, picks out the
property or relation in situations that makes it appropriate or
reasonable to apply the term T even thought the situations in question
do not involve anything to which the term T truly applies, and,
secondly, is such that it is possible to analyze the term T in such a
way that the analysis involves the term S. We have, therefore,
the following conclusion:
5. If a term is definable via weak ostension, then it is verbally definable.
Or, equivalently:
6. If a term is not verbally definable, then it is not definable via weak ostension.
But the following thesis is certainly true:
7. If a term is definable only via ostension, it is not verbally definable.
Theses 6 and 7 then entail:
8. If a term is definable only via ostension, it is not definable via weak ostension.
But the following thesis is also true:
9. If a term is definable only via ostension, it is definable either by strong ostension or via weak ostension.
Theses 8 and 9 then entail:
1. If a term is definable only via ostension, it is definable by strong ostension, but not via weak ostension.
So the first thesis is true.
An Argument against Theses 2 and 3
The objection to thesis 2, and so also to thesis 3, which entails thesis 2, is that there can be terms that correspond to conjunctive concepts. One might, for example, use the term “green-and-round*” as a term that applies to a part of one’s visual field if and only if that part is both qualitatively green and also round. The meaning of that term can be learned via strong ostension, but not via weak ostension. But if one has learned the meanings of the terms “green*” and “round*” via strong ostension, then one can acquire an understanding of the meaning of the term “green-and-round*” verbally by the following definition:
“X is green-and-round*” = def. “X is green* and X is round*”
So thesis 2 is not true: terms that can be learned via strong ostension but not via weak ostension can in some cases be verbally defined.
An Argument for Thesis 4
The argument for the fourth and final thesis requires the following thesis, which I shall not argue for, but which I believe is true:
10. Every term is in principle definable, either verbally or via ostension.
Some terms, however, are not verbally definable. Consider, for example, the term “red*” – the predicate whose meaning is connected with the property of qualitative redness. So we have:
11. Some terms are not verbally definable.
Theses 10 and 11 then entail:
12. Some terms are only definable via ostension.
Recall, however, thesis 1:
1. If a term is definable only via ostension, it is definable by strong ostension, but not via weak ostension.
Theses 12 and 1 then entail:
4. Some terms can be defined only via strong ostension.
2.6 The Main Conclusions
The above lines of thought have suggested certain conclusions concerning what sorts of terms should be taken as the semantical basis for the definition of all other terms.
1. The first conclusion is that the terms that should be taken as semantically basic are those terms that a person cannot come to understand via any verbal definition.
2. A second conclusion is that the following very natural proposal is unsound:
The Immediate Perceivability Criterion
A concept is semantically basic if and only if the concept enters into a proposition that can be the content of an immediate perceptual belief.
3. A third conclusion is that the Immediate Perceivability Criterion
needs to be replaced by one of the following two alternatives:
C. The Direct Awareness
Criterion
A concept is semantically basic if and only if the concept applies
to something by virtue of that thing’s having some property such that
one can be directly aware of instances of that property (and similarly
for relations).
D. The Direct Acquaintance
Criterion
A concept is semantically basic if and only if the concept applies
to something by virtue of that thing’s having some property such that
one can be directly acquainted with instances of that property (and
similarly for relations).
The Direct Acquaintance Criterion
A concept is semantically basic if and only if the concept applies to something by virtue of that thing’s having some property that is such that one can be directly acquainted with instances of that property (and similarly for relations).
4. A fourth conclusion is that terms that are semantically basic can only be defined via strong ostension.
<> Doing so is relatively
simple. Thus, though I favor a somewhat deeper account, it is
natural to say that the concept of the future is the concept of what is
later than the present. But many positions in philosophy of time
cannot offer such an answer. One that cannot is presentism, where
this is the view that only presently existing states of affairs
exist. Other approaches that also cannot do so are tensed
views that attempt to analyze the later than relation in terms of the
concepts of pastness, presentness, and futurity, for then defining the
future as what is later than the present would land one in a circle.
A second example is from
ethics. G. E. Moore maintained that there were “non-natural”
properties of goodness and badness of which one is directly aware, and
he also held that the terms “good” and “bad” could not be
defined. But if that were right, then how could J. L.
Mackie’s error theory be true, for it asserts that there are no
objective values. So it is only possible for Mackie’s error
theory to be true if normative terms are not analytically basic.
But if normative terms are not
analytically basic, how can then be defined? If one defines them
in terms of non-normative, descriptive terms, won’t that imply that
normative states of affairs analytically supervene upon non-normative
ones – a view that many ethical theorists certainly want to reject?
The answer is that in the 20th
century a method was discovered for defining theoretical terms, and one
that does not entail that theoretical states of affairs analytically
supervene upon non-theoretical states of affairs. Consequently,
one can use that method to define normative concepts in such a way that
the truth of normative propositions does not logically supervene upon
non-normative facts.