MAJ __.- w“-.. _‘_. Anw“ A. M“.—
' .
.. . Q I. - . .

SENSE, CERTAINTY AND SIGNALING

Thesis for the Degree cI M. A.
MICHIGAN STATE UNIVERSITY
I Roy I. Gift
1960

THESIS

LIBRARY

Michigan Sta.
University

 

SENSE, CERTAINTY AND SIGNALING

BY
Roy I. Gift
A THESIS

Submitted to the College of Science and Arts of Michigan
State University of Agriculture and Applied Science
in partial fulfillment of the requirements
for the degree of
MASTER OF ARTS

Department of Philosophy

1960

ACKNOWLEDGEMENT

I should like to thank Robert Jones for
helpful suggestions. I cannot sufficiently
express my appreciation nor at all measure
my indebtedness to Professor Richard Rudner
for his encouragement and help in the preparation
of this thesis as well as in so many other

matters during the past few years,

Chapter
Chapter

Chapter

I

II.

III

Chapter IV.

Chapter

V

CONTENTS

CHAPTER I
INTRODUCTION

The purpose of this thesis is to point out a little
(i,e, not sufficiently) noticed relation between the
problems of defining such troublesome semantical predicates
as "true", "denotes" and "means" and some problems of
epistemology, particularly those concerning the grounding
of empirical beliefs.

At the outset, establishing a connection may not
seem too problematical, Yet, the problems of semantics
are not in any obvious sense problems concerning the
justification of beliefs, Rather, they result from an
attempt to give an adequate analysis of semantical
predicates -- and these latter, as we shall see in this
thesis constitute a somewhat different kettle of fish,

I

Epistemological problems, prima facie, are problems

 

 

of justification, Allowing, even, that insofar as these
are philosophical problems what is demanded here, as well,
is an adequate analysis of troublesome predicates, one
still should note that epistemologists are concerned with
predicates such as "credible", "knows", "believes", etc,

which are pragmatic rather than semantical predicates.

 

In each case, the problems arising are many and

familiar. They will not be reviewed here.

Instead, let us recall one aspect of each of these
sets of problems. It is required for the analysis of seman-
tical predicates that proffered definitions be relativized
to a specified language.1 This requirement results from the
possibility of a sentence that is true in one language

being false or meaningless in another, A rule 2: truth

 

for a specified language asserts that a sentence of the
language will be called "true" on the condition that a
semantical relation specified by the rule obtains between
the sentence and extralinguistic events.

As epistemologists, suppose we are concerned with
rules of confirmation for scientific theories. We would
like to present an adequate analysis of the predicates
"confirmable" and verifiable". We assert that a theory
is confirmed when certain logical relations obtain between
the theory and a batch of Eggs observation statements. Of
course, no theory is ever completely confirmed. But that
is not the issue here. The point here is to call attention
to the semantical problem of defining "true" for the obser;
vation statements of a theory or for some statements which
will confirm observation statements.

C. I. Lewiszhas proposed that there are some statements
that are (and indeed must be) immediately verifiable. The

meaning and truth of these statements is indubitably

 

recognized in experience, He takes the position that such

3
statements are necessary for confinnation of theories.
Adopting a slightly worn usage we shall call such statements
"protocol statements".

Without protocol statements, according to Lewis, fig can
only establish logical relations among the statements 2:
science, and "no logical relationship, by itself, can ever
be sufficient to establish the truth, or the credibility
even,,of any synthetic judgment, Unless the beliefs so
related, or some of then, have a warrant which no logical
principle can assure, no logical relation of them to one
another constitutes a scintilla of evidence that they are
3
even probable."
Goodman disagrees with the view that these problems of
confirmation require the certainty of some statements
for their solution. Basically, I take his position to be
that Lewis, without explicit awareness, is really concerned
with two, closely related problems, the confounding of
which has tended to mislead him.
In discussing Lewis's position, Goodman says:
Consider the familiar problem faced by a
common version of pragmatism. The meaning and
truth of a statement are said to lie in its
predictive consequences. These consequences
are themselves statements; for from statements
we can deduce, or even infer with probability,
nothing but other statements, But, if the
truth of these predic tions depends in turn upon
the truth of others derived from them, we are
lost in an endless regress. The theory rests

upon our being able, when a particular moment
arrives, to decide with same degree of confidence

whether a predic tion made concerning that
moment is or is notfulfilled, Addordingly,
statements describing immediate experience

are specifically exempted from the predictive
criterion. But what, then, is to be said
concerning them? What sort of relationship to
experiencg makes such a statement true or
credible?

There are two problems raised in the final question
of this passage. What sort of relationship to experience
makes such a statement true? And what sort of relationship
to experience makes such a statement credible? The first
of these is a semantical problem. The second is a prag-
matic or epistemological problem.

Goodman's proposal for solution of the semantical
problan, one Which establishes the relationship of the
latter to pragmatics in epistemology, is that - semantical
predicates. such as "true", "denotes" and "means"
be systematized in terms of some non-semantical primitive.
He conjectures that a predicate available to be accepted
for the systematization of s semantical predicates is "is
a genuine signal of". Semantical predicates should turn
out to be definable in terms of this primitive, if the
systematization is to be adequate. It should be noticed
that this proposed analysis of semantical .predicates
would subsume them under a more general relation: a

relation obtaining between or among extralinguistic events

EE well gé between linguistic and extralinguistic events.

 

It should be obvious that this does not at all

preclude the possibility of analysing "credibility",

or the justifying grounds for asserting an empirical
judgment, in terms of the grounds justifying taking it
as a genuine signal (or taking the statement of it as
a genuine signal). This ii an epistemological problem,
and I shall maintain that not only is this possibility
not precluded but that, in point of fact, it 33 this
problem to which Goodman is addressing himself under

the rubric of the "The Problem of Projectibility" in

 

.2125: Fiction and Forecast.

To Summarize Goodman's proposals, the important
points are the following:

1) Semantical predicates may be definable in terms
of the predicate "is a genuine signal of".

2) Successful projection of a hypothesis is, under

 

specified circumstances, an index as to its genuineness
as a signal. Anticipating considerations to be presented
below, we might say that, roughly, under specified
circumstances one may be justified in accepting as a
genuine signal a hypothesis containing only well-entrenched
predicates. (An outline of the details of these
proposals is presented in Chapters 4 and 5 of this thesis.
The first of these proposals is put forth'Sense and
Certainty", the second in Fact, Fiction and Forecast.

II

Lewis has been quite critical of"sense and Certainty"

The following is a crucial passage from his reply to
Goodman:

In order to be knowledge, empirical judgment must
not only have predictive import of what will
verify or confirm it; it must also be distin-
guished from a merely lucky or unlucky guess

or hazard of belief by having some justifying
ground. And in the nature of the case, what so
justifies an empirical judgment cannot be sane-
thing future to it and presently uninspectible
but must lie in something antecedent to or
compresent with it. Where it is perceptual
cognition which is in question, the point is
that the interpretation of experience-~the per-
ceptual belief--is significant of the future and
verifiable, but,—in order that this belief have
validity, that which functions as thegnnund of
it muc be present and given.

...And it is on account of that point that I
could not accept Goodman's pragmatic proposal:
by interpreting empirical findings in terms

of what is future to them, it would invite
confusion of the ground of knowledge which is
there and gigen with what is not there but
anticipated.

But to take these remarks as a cogent criticism of
"Sense and Certainty" is to confuse point (1) above
with point (2). Though it is the case that the justify-

ing ground for believing an empirical judgment must lie

 

in conditions antecedent to or compresent with it, it

is precisely the problem of Which antecedent conditions are
relevant to belief that Goodman considers in Eggt, Fiction
and Forecast. That he does not concern himself with this
problem in "Sense and Certainty" probably results from the
fact that there he is concerned with proposal (1) and

not with the problems of validation or justification.

III

Now, though these remarks have touched upon the
relation of importance we hoped to indicate between
semantics and epistemology, we have not clearly in-
dicated what this relation is. On one hand, there is
the problem of systematizing the predicates of semantics;
and, on the other, there is the problem of presenting
adequate theories of confirmation, lawlikeness, and
projectibility, required for systematization of the
predicates of science. In what sense are these dependent
problems?

The following comments appear in Fact, Fiction and

 

Forecast:

If I am at all correct, then,the roots of inductive
validity are to be found in our use of language.

A valid prediction is, admittedly, one that is in
agreement with past regularities in what has been
observed; but the difficulty has always been to
say what consitutes such agreement. Thesnggestion
I have been developing here is that such agreement
with regularities in what has been observed is a
function of our linguistic practices. Thus the
line between valid and invalid predictions (or
inductions or projections) is drawn upon the basis
of how the world is and has been described and
anticipated in words.

The dependence of the problems, then, is somewhat as
follows: an adequate theory of cbnfirmation requires the
distinction of lawlike from non-lawlike statements.

This, in turn, requires a theory of projectibility, which
"is drawn upon the basis of how the world is and has been

described and anti cipated in words." But under what

8
circumstances has the world been described and anticipated
in words?

A concern with this latter problem i§.§ concern with

 

the "root relations" between lagguage and that which it

 

 

 

describes; and this is the problem discussed in "Sense

and Certainty". To say that some bit of language describes
or anticipates an event is to say, in the terms of "Sense
and Certainty", that this bit of language is in a relation-
ship to the extralinguistic world (e.g."true"), which
relation is definable by the relation "is a genuine signal
of".

Thus, the general problem is with the definability of
both the semantical and pragmatic predicates mentioned in
terms of a relation which is pragmatic; i.e. which holds
among extralinguistic events as well as among linguistic
and extralinguistic events.

If Goodman has shown this definability, then he
has shown how such predicates as "true", "means",
"credible" and "is a well entrenched predicate" are
reducible to a common basis. This would amount to an
extremely important step in providing a systematized
language of the predicates of science.

The central problem of this thesis, then, is to
expand and clarify the none-too-clear glimpses of

arguments thus far presented and to scrutinize the

9
adequacy of Goodman's position when compared with some

alternative views.

NOTES

[Numbers following author's name refer to
bibliography.]

l. Tarski (16), p. 14..
2. Lewis (7), (8), (9).
. Lewis (8), p. 169.

. Goodman (5), p. 164.
5. Goodman (4),

6. Lewis (8), p. 175.

7. Goodman (4), p. 117.

10

entries in

11
CHAPTER II
In the introduction it was indicated that we are
going to be concerned at least in part with an epistemolog-
ical problem, and some attempt was made initially to
characterize this problem. In addition, we hinted that
one of the positions alternative to Goodman's on this
problem, and one to be considered, is that propounded by
Lewis. By focusing attention on the pap:e rs presented by
Lewis and Goodman in a symposium in which they both
participated, we shall be able to characterize more
fully the problem and to indicate the pivotal differences
between their points of view. Throughout it should be
kept in mind that our end in view is to present a
common basis for some important semantical and pragmatic
predicates, and our comments are partly direc ted toward
this end.
I
In "The Given Element in Empirical Knowledge"1 Lewis
offers an effective summary of his conclusions on the
grounding of empirical knowledge as follows:
...I seem to find only two alternatives for
a plausible account of knowledge: either there
must be some ground in experience, some
factuality it directly affords, which plays
an indispensable part in the validation of
empirical beliefs, or what determines empirical
truth is merely some logical relationship of
a candidate-belief with other beliefs which

have been accepted.2

These two alternatives are, Lewis believes, exhaustive.

12

His strategy of argument with Goodman is roughly the
dilemma complicated by what appears to be an interpreta-
tion of "or" in the exclusive sense. In his argument,

he attempts to show that Goodman denies the first of the
above alternatives and is, therefore, committed to an
objectionable coherence theory of truth plus abandonment
of the traditional sense of "empirical knowledge".

In order to determine if Lewis is correct in this
charge, it will be necessary for us to examine the first
alternative with some care. In particular, we should
scrutinize Lewis' position to determine what sort of
"factuality" he takes to be directly afforded in exper-
ience. A key to his position is the meaning of "directly
afforded". Of course, also crucial to his argument is
the indispensability of this directly afforded factuality-

I I take it that Lewis' position is a familiar one.
Only its familiarity justifies the brevity of the account
to be presented here, for my eventual aims do not permit
the detailed, lengthy exposition otherwise merited. Let
the following serve, perhaps, as a reminder.

As a beginning, let us recall a passage quoted in the
introduction. We noted there that Lewis says:

Where it is perceptual cognition which is in
question, the point is that the interpretation of
experience--the perceptual belief--is significant
of the future and verifiable, but, in order that

this belief have validity, that which functions
fig the ggound’gfliy must 23 present and giVen.3

 

 

 

13
For Lewis, then, there are three basic aspects of
empirical knowledge: first, perceptual cognition or
empirical judgment and its significance; second,
perceptual belief; third, that, present and given, which
functions as the ground of perceptual belief. Perceptual
beliefs are predictive, whereas, we shall see, that which
is present and given is not. We must now ask what it is
that is directly afforded in experience, i.e. present and
given.

For Lewis, we describe the presentational element of
experience when we use the language of appearances. This,
as he calls it, is language in its "expressive use".4
When I say, for example,5 "There is a flight of granite
steps before me," I commit myself to something beyond
my present experience, according to Lewis. I am committed
to the existence of the steps, at least. But a statement
in phenomenal language, a protocol statement, involves:

no such commitment. "I see what looks like granite steps

 

before me" describes only what is given in my experience.
As Lewis puts it:

The distinctive character of expressive language,
or the expressive use of language, is that such
language signifies appearances. And in thus
referring to appearances, or affirming what
appears, such expressive language neither asserts
any objective reality of what appears nor denies
Eyy. It is confined t3_description of the content
0 presentation itself.

 

 

 

 

14

According to Lewis, this argument supports two
contentions of his. This lack of commitment to "objective
reality" is construed by him to indicate that protocol
statements are both non-predictive and incorrigible. If
what a protocol statement describes is confined to an
evanescent moment of experience, then it says nothing
about future experience; hence, it is non-predictive. And
after that moment has vanished, therefore, there is no
experience relevant to the disconfirmation of this protocol
statement; hence, it is incorrigible.

Another argument that Lewis presents for the incorrigi-
bility of protocol statements was mentioned in the intro-
duction. There we pointed out that Lewis takes indubitable
protocol statements to be necessary for the confirmation
of theories.

Lewis makes a distinction between terminating and
non-terminating judgments, on which this new argument
relies. Briefly, a terminating judgment is rendered in
expressive language. For instance: "If I feel that I
am walking toward what looks like a flight of granite steps,
then each step will appear to become larger" approximates
a terminating judgment. For Lewis, terminating judgments
are characterized by being completely confirmable or
disconfirmable in experience. Non-terminating judgments,

on the other hand, such as statements of science, are

15
never completely confirmed. Continuing to discuss
non-terminating judgments, Lewis says:

However, if the suggested account should be
correct, then the judgment of objective fact implies
nothing which is not theoretically verifiable.
And since any, even partial, verification could
be made only by something disclosed in some
passage of experience, such an objective and
non-terminating judgment must be translatable
into judgments of the terminating kind. Only

so could confirmation of it in experience come
about. If particular experiences should not
serve as its corroborations, then it cannot

be confirmed at all; experience in general would
be irrelevant to its truth or falsity; and it
must be either analytic or meaningless.

The argument runs, then somewhat as follows:
requisite to the confirmation of non-terminating judgments
is their translation into terminating judgments. Though
terminating judgments can result in incorrigibility, they
are subject to error and, hence, qualify as knowledge
for Lewis. Still, non-terminating judgments are never

completely confirmed, because they translate into an
infinite set of terminating judgments. On the other hand,
to say that protocol statements are dubitable would imply,
for Lewis, that we are unable to recognize a such and so
experience 2g such and so when we have it. (I am assuming
that to recognize a such an so experience is to identify
it as that sort of experience. Since Lewis asserts that
mistakes in labeling can be made, I suppose he denies that
identifying and labeling come to the same thing.)8 He

maintains that if we cannot recognize an experience for what

16
it is when we have it, then we cannot verify terminating
judgments and, hence, cannot confirm non-terminating
judgments. Thus, according to Lewis, the indubitability of
protocol statements is a necessary condition for the
confirmation of theories, i.e. non-terminating judgments.

Furthermore, he maintains, without indubitable
protocol statements theories of science lose the
connection to experience required by empiricism. Without
this "connection to experience" we can define truth for
the statements of science only in terms of relationships
holding among such statements. This amounts to defining
truth in terms of logical relations among statements and
is, thus, a fonm of(bbjectionable) coherence theory of
truth.

One further comment is required before proceeding
with our exposition of Goodman's "Sense and Certainty".
Throughout our exposition, it seemed we were discussing
the truth of terminating and non-terminating judgments as
viewed by Lewis. But now we come upon an even more
interesting problem. Non-terminating judgments are never
known to be true, i.e. never completely confirmed. We
can, at most, make a decision as to their credibility.

According to Lewis, this also depends 22 protocol state-

 

ments and their indubitability.

 

In order to better illustrate the way in which this

l7
situation becomes problematic, let us turn to an account
of Goodman's position.
II
It is, perhaps, best to begin by summarizing the
points we made in the introduction. Goodman takes Lewis
to have confused two problems, expressed by the question:
What sort of relationship to experience makes a statement
describing immediate experience true? And what sort of
relationship to experience makes such a statement credible?
The first of these, Goodman proposes, may be solved
through defining truth in terms of signaling. To illustrate
how such a definition might proceed, let us suppose a
statement "5" says there will be a red patch in the visual
field at some future time, 3. Goodman continues with the
example as follows:
Suppose the statement 2 "There is now a red patch
in the visual field" occurs at the time 3 above in
question. 3 does not sional the simultaneous
occurrence of the red pafch; for signaling is
always forecasting. Nevertheless, we know that
g is true if an only if‘fi is true. Hence 3 is
true just in case E is a genuine signal. Although
g does not itself signal the occurance of the
red patch, the truth of P is explained in terms
of the truth of the earlier statement 5, which
does signal this occurence. Statements in the
past tense can be handled in the same way as those
in the present tense; and tenseless statements,
depending on whether they occur before, during, or
after the time they pertain to, in the same way
as statements in, respectively, the future, present,
and past tense.

This is obviously not a completed definition and there

18
are problems yet to be solved. Some of these we shall
uncover in the next chapter, some in chapter 5. Yet, this
passage does outline a strategy for giving such a definition,
and a crucial task of this thesis will be an assessment of
whether this strategy will work.

If this strategy does work, then one among its
advantages will be our ability cogently to reject the thesis
of the indubitability of protocol statements and thus "save"
a major tenet of post—Kantian empiricism: namely, that all
non-analytic statements are corrigible. We will have
adequate means for accounting for the truth of terminating
and non-terminating judgments on the basis of their pre-
dictiveness; and we will be able to define "true" for pro-
tocol statements, while admitting that there is, and
explicating some sense in which they are non-predictive.
Such a definition would show there is 22 theoretical
requirement in empiricism that protocol statements be
indubitable.

Though Goodman has rejected the first alternative in.
the dilemma, to the extent that it requires the indubita-
bility of protocol statements, we shall see that he is,
nevertheless, not committed to a coherence theory of truth.
On the contrary, the signaling relation is a broad basis

including a wide variety of experiences, and his theory of

 

projectibility calls for a variety of data as to the

l9
predictive use of language and verificatory experiences.
However, as we pointed out earlier, Lewis seems also
to take the indubitably present and given to be required
for the credibility even of terminating and non-terminating
judgments. In the passage with which we began this exposi-
tion, Lewis says:

...the interpretation of experience-~the perceptual
belief--is significant of the future and veri-
fiable, BHt, in order that this belief have validityL
that which functions as the ground of it must Be
present and given.

...And it is on account of that point that I
could not accept Goodman's pragmatic proposal:
by interpretating empirical findings in terms of
what is future to them, it would invite confusion
of the ground of knowledge which is there and
given with what is not there but anticipated.10

 

This judgment that Goodman's proposal is too pragmatic
results from Lewis' confusion, mentioned above, of the
purely semantical with the epistemological problem. He
oonstrues Goodman as saying we must await the occurence
of g before we can decide upon the credibility (or as
he puts it "validity") of F. But it is simply not the case
that Goodman says this. The occurrence of P (and other
events as well) determines the.££2£fl of F, but we can
decide F's credibility on quite other grounds, e.g. the
precise grounds that Goodman adduces in terms of pggg
predictions in his theory of projectibility.

The theory of projectibility in Fact, Fiction and

 

Forecast is addressed to the problem of determinégthe
projectibility of generalized hypotheses, rather than

to statements such as'f or P_above. Yet, the theory can

20
be readily extended to cover singular hypotheses.

There are two reasons why this extension is virtually
automatic. First, projectibility of a general hypothesis

is inherited by its instances. Hence, if we are concerned

 

 

to determine the projectibility of some singular hypothesis,
we can do so by appropriately generalizing this hypothesis
and determining the projectibility of the hypothesis of
which the one in question is an instance. Second, it is

a logical thesis of both Goodman and Quine that any singular
statement is equivalent to some general statement; thus,

comments in Fact, Fiction and Forecast are applicable to

 

 

singular statements after all.

Though Lewis is right in pointing out the pragmatic,
predictive aspect of Goodman's account, he is incorrect
in assuming credibility must be decided 223 on the basis
of the future outcome of present predictions. This would
be an odd situation indeed, but it is not entailed by
Goodman's proposal.

On Goodman's view, prima facie credibility amounts
to prima facie projectibility. It is not required that
a hypothesis be true, only that the total relevant
information available warrants use of the hypothesis for
predicting.

Thus, we may conclude that what is required is a
clear notion of the genuine signal relation and an
exposition of Goodman's theory of projectibility understood
as a reply to Lewis, i.e. as an attempt to provide grounds

of credibility contemporaneous with the need for them.

21

In the meantime we shall examine the semantical
definition of truth. Throughout this chapter we have
referred to the (as yet here) unclarified notion of
truth. It has sometimes been supposed that an adequate
definition of "truth" would resolve some of the difficulties
we have encountered.11 Thus, we turn to a definition that
has been proclaimed adequate for epistemology to discover

if this is the case.

45 La.) N H
o

00 \I 0‘ (J!
o

10.
11.

22

NOTES

 

C. 1. Lewis (8).
Ibid., p. 168.
Ibid., p. 175, my ital.

See, for example, exposition of this use of language
in (7), pppo 178-826

This example is taken from (7), p. 179.
Ibid., p. 179.
Ibid., p. 181.

For significance of this parenthetical remark, see
(6), pp. 99-101, and (5), pp. 160-1.

Goodman (5), p. 166.
Lewis (8), p. 175.

See, for instance, A. J. Ayer (l), p. 238.

23
CHAPTER III

In the introduction it was pointed out that we are
concerned to indicate an important relation between prob—
lems of defining semantical predicates, e.g. "true", and
problems of defining pragmatic predicates, e.g. "credible".
At the end of Chapter II, we noted that some philosophers
have assumed that an adequate definition of truth will
solve the problem of defining "credible". Since Tarski's

1 has been proclaimed adequate by

definition of truth
some, we shall examine his proposals in order to determine
if this is indeed the case.

There are three points important for us to consider:
1) Tarski's definition is given only for certain, pre—
cisely specified, formal languages.
2) The definition of truth is given in terms of "satisfies",
which Tarski takes to be 5 semantical predicate, but
which others, e.g. Bergmann, have taken to be a syntactical
predicate, as used by Tarski.
3) In the light of the above we must scrutinize the
adequacy of Tarski's definition in relation to Goodman's
proposals.

Tarski requires of his explication of the semantical

predicate "true" that his definition of truth be both

materially adequatg and formally correct. Our purpose

 

 

in this chapter will be to question the adequacy of Tarski's

definition, taken as a proposed means to solve the problem

of defining "credible". We may find grounds for deter-
mining that his definition is inadequate for this pur-
pose, even though the definition is considered materially
adequate and formally correct according to Tarski's own
explicit criteria.2
I

In order to specify a criterion of material adequacy,
Tarski considers the extension and,intension of the predicate
"true".

In regard to the extension of the term he says:

The predicate "true" is sometimes used to refer
to psychological phenomena such as judgments or
beliefs, sometimes to certain physical objects,
namely, linguistic expressions and specifically
sentences, and sometimes to certain ideal entities
called "propositions." By "sentence" we understand
here what is usually meantfi in grammar by "declar—
ative sentence"; as regards the term "proposition,"
its meaning is notoriously a subject of lengthy
disputations by various philosophers and logicians,
and it seems never to have been made quite clear
and unambiguous. For several reasons it appears
most convenient to apply the term "true" to
sentences, and we shalI'follow this course.5

 

 

 

He adds in a footnote that he understands "sentences" to
refer to "classes of inscriptions of similar form"4, to
classes of physical things rather than to individual
physical things. Put another way, "true" is taken by
Tarski to be a predicate of sentence types.

This results.in a limitation on the definition of

truth. The definition must be relativized to a particular

language, according to Tarski, since different languages

25
may contain sentence tokens of the same type with differ-
ing truth values. Thus, for Tarski, "true" is a predicate
of sentence types that must be defined in regard to the
tokens of a type appearing in some language or other.
Tarski gives the criterion of material adequacy as
an intensionalistic criterion. He wants his definition
to conform to the actual meaning of the term "true" as
used classically.
Consider the sentence "snow is white". We ask
the questhan under what conditions this sentence
is true or false. It seems clear that if we
base ourselves on the classical conception of
truth, we shall say that the sentence is true
if snow is white, and that it is false if
snow is not white. Thus, if the definition of truth

is to conform to our conception, it must imply the
following equivalence:

 

The sentence "snow is white" is true if, and only

 

if, snow is white.7

 

This procedure may be generaliZed as follows: first,
we arbitrarily Choose a sentence and replace it by the
letter 'p'. Next, we replace the name of this sentence
by the letter 'X'. Then, an equivalence of the form we
are concerned with is written:

X is true if, and only ifJ p.

 

"We shall call any such equivalence (with 'p' replaced by
any sentence of the language to which the word "true" refers,

and 'X' replaced by a name of this sentence) an 'gquivalence

 

of the form (T).'"6

 

Tarski's criterion of material adequacy may be
phrased as follows: A definition of truth is adequate

only if it implies all equivalences of the form (T).

26
There are two points to be considered in regard to
formal correctness. First, the language for which "true"
is to be defined must be exactly specified. This is
required to avoid the occurrence of antinomies which
would vitiate the definitions. Tarski uses the antinomy
of the liar as an example which involves the predicate
"true". Though we shall not present this antinomy here?
we shall discuss the assumptions which lead to it.
According to Tarski, these are:
(I) We have implicitly assumed that the language
in which the antinomy is constructed contains,
in addition to its expressions, also the names of
these expressions, as well as semantic terms such
as the term "true"referring to sentences of this
language; we have also assumed that all sentences
which determine the adequate usage of this term
can be asserted in the language. A language with

these properties will be called "semantically
closed."

 

(II) We have assumed that ig this language the
ordinary laws of logic hold.

Tarski considers the consequences of rejecting
assumption (II) to be too obvious to require elaboration.
He concludes we must avoid using a language that is semantical-
ly closed.

This brings us to the second point. Tarski takes the
problem of formulating a semantical metalanguage for a
natural language to be a vague,and therefore, invalid prob-
lem. In natural languages we do not know precisely what

are the sentences and which sentences are assertible. It

27

is difficult to decide, for instance, whether or not
natural languages are semantically closed, in the sense
specified.9

The metalanguage Tarski has in mind must contain all
of the sentences of the object language. It must be
rich enough for the construction of names for these
sentences, and it must have a basic logical apparatus.
Tarski also points out that the conditi>n of essential
richness of the metalanguage, which is not to be further
explored here, is both a necessary and sufficient
condfiion for a satisfactory construction of the definition
of truth.

In addition to all these requirements, Tarski says:

It is desirable for the metalanguage not to contain
any undefined terms except such as are involved
explicitly or implicitly in the remarks above, i.e.:
terms referring to the form of the expressions

of the object-language, and used in building names
for these expressions; and terms of logic. In
particular, we desire semantic terms (referring to
the object-language) to be introduced into the
meta-language only by_definition. For, if this
postulate is satisfied, the definition of truth,

or of any other semantic concept, will fulfill what
we intuitively expect from every definition; that
is, it will explain the meaning of the term being
defined in terms whose meaning appears to be
completely clear and unequivocal. And, moreover,
we have then a kind of guarantee that the use of
semantic concepts will not involve us in any
contradictions.

 

 

 

II
The definition of truth is given in terms of the

notion of satisfaction. Satisfaction is defined through a

28

recursive procedure; and, as it turns out, "a sentence
iseither satisfied by all objects, or by no objects.
Hence we arrive at a definition of truth and falsehood

iéfififl 21 31$ objects, and false otherwise."11

 

 

According to Tarski, this definition is materially
adequate; and, in addition, since the definition implies
all equivalences of the form (T), it uniquely determines
the extension of the term "true". (Thus, any other
materially adequate definition of truth would be equivalent
to, or "imply", the one given by Tarski.

If we agree that the definition is materially adequate
and formally correct as Tarski claims, we may still raise

other questions of adequacy. Bergmann,12 for instance,.

 

has argued that Tarski has not defined the semantical
predicate "true", since "satisfied" in terms of which
"true" is defined is not, as Tarski maintains, a semantical
predicate but is, rather, a syntactical predicate. Thus
Bergmann concludes, Tarski has defined a sense of "true"
other than the philosophically important, semantical
predicate "true".

Yet even if we assume that Bergmann is correct in

taking Tarski's predicate "satisfied" to be a syntactical
predicate, it does not follow that Tarski's predicate

"true" is not a semantical predicate. It may mean, instead,

29
that Tarski has broken the semantical circle through
defining a semantical predicate in terms of a non-
semantical predicate. Fortunately, we need not attempt
to decide this issue here. It is sufficient for our
purposes to note that if Tarski has broken the semantical
circle, he has done so with a syntactical predicate rather
than with a pragmatic one.

We are, after all, concerned with defining "credible"
for phenomena . If Bergmann is correct in maintaining
that "satisfied" as used by Tarski is a syntactical
predicate, his argument fortifies the claim that an
adequate definition of "credible" cannot be given in
terms of Tarski's sense of "true".

Tarski attempted to utilize the clear specification
of truth conditions given by his definition to resolve
epistemological difficulties. One such attempt is a
postulate he proposed on the acceptability of scientific
theories. This postulate reads:

As soon as we succeed in showing that an
empirical theory contains‘(or implieS) false

sentencesL i: cannot be any longer considered
acceptable.15

 

 

 

 

Yet this is a useful postulate only when we have
means available for determining when a sentence of a
theory is false. Tarski indicates such means for theories

that are self-inconsistent. I am sure there are many

3O
arguments for the unacceptability of an inconsistent
theory. But how do we know a theory is no longer
acceptable when it is self-consistent or, at least,
when we are unable to prove it inconsistent? In short,
the general problem of acceptability is one of decisx>n
in the absence of definite information as to the truth
or falsity of any of the statements of the theory. The
semantical conception of truth does not provide a
solution to this problem.

Though the definition does define truth conditions
for the statements of a theory, it does not provide
solutions to the varinus problems of confirmation. An
adequate analysis of confirmation will employ the
semantical prodicate "true"; yet, here again, a problem
in defining degree of confirmation is that such be
assignable in the absence of definite information as to

the truth or falsity of any sentences of the theory.

The pragmatic predicates of epistemology are not fully

 

definable in terms of the predicate, "true", as defined

by Tarski. The semantical definition of truth gives the
condifions under which a sentence of a theory is true;

but it does not tell us how we, as fallible observers, can
decide when this condition is fulfilled. Obviously, it

is asking too much to demand this of Tarski's analysis.

31

Yet, we cannot dismiss the issue by saying that an
adequate analysis of epistemological predicates will
employ the semantical predicate "true". We should like to
show how this predicate can be employed in such an
analysis. In the next chapter we shall examine the
question of whether or not Goodman's theory of projectibility,
supplemented by the semantical predicate "true", is

adequate to define "credible" for phenomenal statements.

10.
ll.
12.

13.

32

NOTES

 

Tarski (16).

We accept Rudner's view-(l4), p. 15--that the
adequacy of an analysis is to be judged by the
efficiency of the analysans in serving most
important purposes served by analysandum.

 

 

Tarski (16% p. 14.
Ibid.,p. 42.
Ibid.,p. l6
Ibid.,P. 16
Ibid.,pp. 19-20.
Ibid.,p. 20.

Max Black has taken issue with Tarski on the

invalidity of this problem. He points out means of
clarifying and solving the problem in such a way

that Tarski's definitions may be rendered applicable

to a natural language, Yet,though this problem is
interesting and pertinent to our own problem, consider-
ing it in detail here would amount to a lengthy
digression that we cannot afford, however tempting.

Tarski, Op.Cit., pp. 22—3.

Ibid p. 25.

0’

Bergmann (2), pp. 17-23.

Tarski, Op. Cit., p. 39.

33
CHAPTER IV

It was pointed out in the introduction that an
adequate theory of confirmation requires a theory of pro-
jectibility. In this chapter, something of the rationale
for this claim will be presented in a discussion of Good-
man‘s theory of projectibility.

According to Goodman, the theory of projectibility
"is drawn upon the basis of how the world is and has been
described and anticipated in words." Our task here will be
specifically to investigate Goodman's proposed theory of
projectibility with a view to establishing how this
dependence of confirmation on projectibility impinges on

the problem of the grounding, or credibility of phenomenal

 

statements. In the course of this investigation we shall
attempt, also, to show how this theory is related to the
semantical problems previously considered. The ultimate

goal will be to show that "projectible" is definable in terms
of the predicate "is a genuine signal".

For our exposition of Goodman’s theory of projectibility
we shall concentrate on two chapters from 3233, Fiction and
Forecast, the chapters entitled "The New Riddle of Induction"
and "Prospects for a Theory of Projection".-1

In "The New Riddle of Induction", Goodman briefly

covers the history of the problem of justifying induction.

He points out that the problem of justifying induction

34
debated in recent years results from a misconstrual
of Hume's treatment of induction.

Hume argued that there are no necessary connections
of matters of fact. His account of prediction is an
explanation of prediction as habit.' We make a prediction
on the basis of its accord with past regularities. This
account has been construed as avoiding the "real" problem of
induction. This problem, according to Hume's critics, is
one of providing a warrant for making predictions. The
counter to Hume's argument is that Hume's account was an
attempt to explain rather than to justify induction. Thus,
the "real"problem of induction was traditionally construed
to require some sort of justification for predictions.

Goodman has, however, given an apt reply to these
criticisms. If the problem is one of justification,
then what sort of justification is required to solve it,
Goodman asks? "If the problem is to explain how we know
that certain predictions will turn out to be correct,
the sufficient answer is that we don't know any such
thing. If the problem is to £229 some way of distinguish-
ing antecedently between true and false predictions, we
are asking for prevision rather than for philosophical
explanation."2

According to Goodman, we can clarify this problem of

justifying induction by calling attention to how deductions

35
are justified. A deduction is justified when it is
shown to accord with the codified rules of deductive
inference. The rules of the codification are justified
insofar as they yield acceptable inferences. But, as
Goodman comments:

This looks flagrantly circular. I have said

that deductive inferences are justified by

their conformity to valid general rules, and

that general rules are justified by their confor-
mity to valid inferences. But this circle is a
virtuous one. The point is that rules and particular
inferences alike are justified by being brought into
agreement with each other. A rule is amended if

it yields an inference we are unwilfing to accept;

an inference is rejected if it viOTates a rule we

are unwiIligg to amend. The process or justification
is the deliCate one of making mutual adjustments
between rules and accepted inferences; and in the
agreement achieved lies the only justification need-
ed for either.

 

 

 

 

If we apply this principle to the problem of justifying
induction, we see that an induction should be justified
by its conformity to validcnnnons of induction. The
cannons themselves are valid insofar as they reflect
accepted inductive practice.

Thus, Hume's attempt to explain actual inductive
practice ESE directly to the point of the problem.
Briefly, HUme's thesis was that confirming instances of
a general hypothesis provide a sanction for using the
hypothesis to predict further instances. His attempt,
however, was unsuccessful, since, as J. S. Mill pointed
out,4 some inductions are incorrect though made on the basis

of a great deal of available evidence. Equally supported

36
hypotheses are not equally sanctioned. Goodman's thesis
that the sanction provided varies with both the support

a hypothesis has and its projectibility.

 

Thus, not all past regularities will support a
prediction. Yet, it is a mistake to constrquume's efforts

as being misguided. The problem requires attention to

 

actual inductive practice in order to define the difference
between valid and invalid predictions.

We have still to indicate the dependence of the
problems with which we introduced this chapter. The prob-
1em of projectibility is one of determining when it is
legitimate to project from any given set of cases to a
wider set. What conditions sanction such an expansx3n?

The problem of induction on the other hand is one of
providing a sanction for projecting from past cases to
future cases. Hence, the problem of projectibility is the
more general problem, and its solution should solve the
problem of induction.

On this latter problem, let us attend to Goodman‘s
comments on the recent efforts of Hempel in confirmathn
theory. The problem of defining yglid induction as approach-
ed by Hempel is one of defining "the relation that obtains
between any statement 31 and another 32 if and only if S1
may properly be said to confirm 82 in any degree."5

As reviewed by Goodman, some progress has been made

in defining this raltion purely syntactically. Various

improvem

criteria

37
ents have been made in the way of tightening the

by which it is determined that which evidence

statements confirm.

Yet

as Goodman points out, satisfaction in this

progress is shortlived. A purely syntactical elaboration

of Hume'

s generalization procedure is subject to criticisms

analogous to those made of Hume's formulation of this

thesis.

'Hempel's criteria also allow generalizatkans

from the available evidence to incredible as well as

credible

hypotheses.6 Other than syntactical features of

a hypothesis are involved in determing its confirmatknn

by an in

stance.

As(3oodman says:

That a given piece of copper conducts electricity
increases the credibility of statements asserting
that other pieces of copper conduct electricity,
and thus confirms the hypothesis that all copper
conducts electricity. But the fact that a given
man now is this room is a third son does not
increase the credibility of statements asserting
that other men now in this room are third sons,

and so does not confirm the hypothesis that all
men now in this room are third sons. Yet in both
cases our hypothesis is a generalizatian of the
evidence statement. The difference is that in the,
former case the hypothesis is a lawlike statement;
while in the latter case, the hypothesrs is a merely
contingent or accidental generality. Only a
statement that is lawlike--regardless of its truth
or falsity or its scientific importance--is capable
of receiving confirmation from an instance of it;
accidental statements are not. Plainly, then, we
must look for a way of distinguishing lawlike from
accidental statements.7

Goodman gives an example to indicate both the

difficulty and the importance of making this distinction.

38

Suppose that for every emerald examined before time
E we can assert evidence statements that each of these
emeralds is green, thus supporting the hypothesis that
all emeralds are green. Then, suppose we introduce another
predicate, "grue", that applies to anything examined before
time i that is green and also to anything else that is
blue. We can then assert in an evidence statement, for
each emerald examined before time £,that it is grue, thus
supporting the hypothesis that all emeralds are grue and
the prediction that the next emerald examined after time
1 will be blue.

There are two important points.illustrated by this
example. First, we can choose other predicates like
"grue" such that our observations of emeralds will provide
support for any predicthan whatever about future cases.
Second, though we £331 we know which of the hypotheses
leading to incompatible predictions is "genuinely support-
ed".by the evidence, we have no criterion for determining
this, i.e. we have no criterion for determining which of
the hypotheses is lawlike.8 Thus, without a criterion of
alawlikeness our generalization criterion of confirmation
is rendered otiose, since it fails to exclude any pre-
dictions whatsoever.

Goodman reviews four proposals for solving the

problem of distinguishing between lawlike and accidental

39
hypotheses, each of which has proved unsuccessful. It
would not be of particular advantage here to consider each
of these proposals. It is enough for our purposes simply
to observe that Goodman's appraisal of each as unsatisfactory
seems conclusive, and shall be so assumed here. According-
ly, we are confronted with the unsolved problem of dis-
tinguishing between accidental and lawlike hypotheses,

"the new riddle of induction", as Goodman calls it.

The remainder of the chapter on the new riddle of
induction is employed in indicating how this problem is
reducible to the problem of projectibility. Goodman points
out that "the problem of prediction from past to future
cases is but a narrower version of the problem of projecting
from any set of cases to others."9 Thus, the problem of
distinguishing between accidental and lawlike hypotheses
is more generally the problem of distinguishing between
non—projectible and projectible hypotheses. But before
continuing with the exposition of Goodman's attack on this
problem, it is necessary to review some of his previous
remarks that serve to characterize the problem of pro-
jectibility.

In Chapter II of Fact, Fiction and Forecast, "The

 

Passing of the Possible", it is noted that the general

problem of dispositions is:

 

...the problem of characterizing a relation such
that if the initial manifest predicate "Q" stands

40

in this relation to another manifest predicate

or conjunction of manifest predicates"A", then

"A" may be equated with the dispositional coun-
terpart —- "Q"-ab1e" or "QD" -- of the predicate
"Q". But the question when such a 'causal' con-
nection obtains or how laws are to be distinguish-
ed from accidental truths is an especially per—
plexing one.

In this grim picture, we can find one small note
of comfort. Observe first that solution of the
general problem will not automatically provide
us with a definition for each dispositional predi—
cate; we shall need additional special knowledge
in order to find the auxiliary predicate that
satisfies the general formula -- i.e. that is re-
lated in the requisite way to the initial manifest
predicate. But on the other hand, discovery of
a suitable definition for a givendispositional
predicate need not in all cases wait upon solution
of the general problem. If luck or abundant special
information turns up a manifest predicate "P" that
we are confident coincides in its application with
"flexible", we can use "P" as definiens for "flex-
ible" without inquiring futher about the nature of
its connection with "flexes".10

The problem of dispositions is thus one of projecting
an initial manifest predicate to cover a wider range.
Scheffler has illustrated "projecting to a wider range"
in terms of defining "is playable" on the basis provided
by "is played".1-1

...we may denote as "played" every record which,
at any time, is actually put on the turntable of
some record player of familiar type and, with the
needle in position, produces recognized music or
speech. Aside from these, we may dermate as
"playable" also certain other objects, e.g.,
records accidentally shattered before ever
reaching the turntable. Asked to explain the
latter term, we should naturally try to do so by
reference to the former via recourse to notions
such as capability, possibility or potentiality,
to the use of the subjunctive, or to devices like

41

reduction-sentences. And each of the latter courses
seems to point to a queerness inherant in the
predicate "is playable"; it talks about something
else than what is actually the case in our world.
Consider, however, that everything in the above
sense playable is a record of music—5r speech,
meeting rathEY dEfinite speEifiEETiofis as to form,
shape, and history, and conversely, that every such
record is playable. [my ital.] This means that

"is playable" is definable in terms of those
predicates by means of which such specifications
are stated, without recourse to possibility or the
subjunctive. Assuming that such predicates are
among the primitive observational terms of a given
system S, we now have a situation in which one
predicate ("is playable") is dispositional with
respect to another ("is played") in S, but is

fully definable and non-dispositional with respect
to certain others, also in 8.12

 

 

 

 

 

 

Thus, there is nothing mysterious involved in
projecting to a wider range. The problem is to find some
set of manifest predicates, applicable to all actually
played objects and certain other objects not actually
played, such that the dispositional predicate "is play-
able" can be defined in terms of these. To accomplish
this for some particular dispositional predicate need
not await solution of the general problem of dispositions,
i.e. we need not concernourselves with questions on
'causal' connections obtaining between "is played" and
"is playable".

It is important to notte that the problem of defining
some particular dispositional predicate may not depend upon
the solution of the general problem. The importance of

this possibility not being excluded becomes obvious when

42

we recognize that the problem 9i projection may be

 

 

construed as one of defining the dispositional predicate

 

"projectible" 23 the basis 2f the manifest predicate

 

 

 

"projected". It is a problem Of projecting "projected".

 

The immense saving of effort that is accomplished by
reducing the problems of lawlikeness and dispositionals
to the problem of defining "projectible" should not be
overlooked. What in effect is accomplished is that the
general problem of dispositionals is neatly reduced to the
more manageable problem of defining one dispositional
predicate, the solution of which should also provide means
for distinguishing accidental from lawlike hypotheses. I
do not intend by these comments to give the impression that
I consider this new problem to be a simple one. I do
desire to call attention to ‘its importance.

Goodman speaks in the following manner about this
new problemzl

The reorientation of our problem may be por-
trayed in somewhat more figurative language.
Hume thought of the mind as being set in motion
making predictions by, and in accordance with,
regularities in what it observed. This left him
with the problem of differentiating between the
regularities that do and those that do not thus
set the mind in motion. We, on the contrary,
regard the mind as in motion from the start,
striking out with spontaneous predictions in
dozens of directions, and gradually rectifying
and channeling its predictive processes. We ask
not how predictions come to be made, but how --
granting they are made -- they come to be sorted
out as valid and invalid. Literally, of course,
we are not concerned with describing how the mind

43
works but rather with describing or defining
the distinction it makes between valid and
invalid projections.

For the most concise presentation of Goodman's treat-
ment of projectibility, we should, perhaps, merely offer
at this point his chapter on the theory of projection
within quotation marks and with proper acknowledgment.

But we shall do our best to summarize without disastrous
loss of detail.

At the close of the chapter on the new riddle of in-
duction, Goodman concluded that this problem could not be
solved on the basis of purely semantical or syntactical
grounds. Thus, in the chapter "Prospects for a Theory of
Projection" he introduces a pragmatic predicate to serve
as primitive in the construction of the definition of
"projectible". Goodman summarizes this chapter as follows:

In brief review, the route of definition I have
outlined starts with the explanation of the primi-
tive "is a much better entrenched predicate than"
in terms of actual projections, and runs as follows:
first, definition of presumptive projectibility
by means of the rules of elimination; second, defini-
tion of initial projectibility index for presump-
tively projectible hypotheses; and third, definition
of degree of projectibility in terms of the initial
index as modified by correlative information
through overhypotheses.14
This summary suggests a helpful division for our exposition.
We shall subsume our expository comments under the follow-
ing topics: "The Basis: Elimination", "Presumptive Pro-
jectibility: Expansion", "Degree of Projectibility".

The Basis: Elimination

Since the purpose of Goodman's analysis is to define

44

"projectible" on the basis of the manifest predicate
"projected", the basis of the analysis will be the
hypotheses that may be said to have been actually pro-
jected prior to a given time. The temporal reference is
necessary throughout, since Goodman does not consider
in these lectures the question of defining "projectible"
without temporal qualification.

A hypothesis will be said to be actually projected
if it is explicitly used to predict the outcome of examining

previously unexamined cases and if it is 33 that time neither

 

violated, exhausted, nor unsupported. A general hypothesis
is violated when a negative instance has been discovered,
it is exhausted when there are no unexamined cases re-
maining, and it is supported when at least one positive
instance has been discovered.

The above criterion isolates the class of actually
projected hypotheses. The strategy Goodman employs in
defining the class of projectible hypotheses involves

solution of two problems. First, unprojectible hypotheses

 

that have been actually projected must be eliminated from
the basis. This is accomplished through introducing rules
based on entrenchment. Second, the basis must somehow be

expanded to include projectible hypotheses that have neve£

 

been actually projected.14a

We can eliminate from the basis all hypotheses that
have either been violated or exhausted since projection.

Though this was a fairly simple beginning, elimination of

45
other hypotheses will have to be accomplished in terms of
rather complicated conflicts. This is further compli-
cated by the fact that it is not desirable to eliminate
all conflicts, since some conflicts must await settle-
ment through further experimentation.' We want criteria
that will settle only those conflicts involving questions
of the validity of one or the other of the hypotheses.
Goodman settles such questions with three rules involving
a comparison of entrenchment.

As an illustration of conflicting hypotheses, Goodman
uses the hypotheses containing the predicates "green" and
"grue". You will recall from our previous use of this
example that the hypotheses containing these predicates
resulted in incompatible predictions. The question is:
how are we to eliminate the invalid hypothesis containing
the predicate "grue"?

We may do this be consulting the history of actual
projections of the two predicates. If we do this, we shall
discover that the predicate "green" has been projected
much more frequently than the predicate "grue". We may then
say that the predicate "green" is much better entrenched
than the predicate "grue". Thus we have:

Rule 1: "A projection is to be ruled out if it conflicts
with the projection of a much better entrenched predicate."15

An important feature of entrenchment is that not only
one predicate becomes well entrenched by its frequent pro-

jection (and it is important to emphasize that it is

46

projection and not merely use that accrues entrenchment

 

for a predicate) but also all predicates coextensive

with it. Thus it is the class which the predicate selects

 

 

which becomes entrenched, and "to speak of the entrench-

ment of a predicate is to speak elliptically of the entrench-
ment of the extension of that predicate."16 This feature

of entrenchment has an advantage in regard to unfamiliar
predicates which we should not, of course, like to discard
out of hand. Such a predicate may inherit entrenchment

from the frequent'grojection of a predicate coextensive

with it.

Another point in regard to unfamiliar predicates, it
should be recalled, is that we are net concerned with the
elimination of predicates at all. The problem with which
we are presently concerned is the elimination of objection-
able hypotheses from those actually projected. Care must
be taken, of course, to insure that new predicates are
not eliminated by rules designed to eliminate such DXBE‘
theses.

Goodman indicates that there is available a means
for broadening the basis of reference hypotheses. This
expansion of basis will enable us initially to strengthen
Rule 1. The expansion is effected by means of the predi-
cate "could have been projected". This predicate is ap-
plicable to any actual hypothesis that has not been actually

projected up to the time in question, that is supported,

47
unviolated and unexhausted at that time.17 These addi-
tional hypotheses, of course, have no affect in deter—
mining entrenchment, since this is determined on the basis
of actual projections.

Now we may revise rule 1 as follOws:

Rule la: A projection is to be ruled out if it conflicts
with the projection of a much better entrenched predicate
or conflicts with a much better entrenched predicate that
could have been projected.

Thus extended, this rule now enables us to throw out
objectionable hypotheses that conflict with projectible
hypotheses that have not been actually projected.

The second rule provides an extension of the class of
reference hypotheses to include some hypotheses that could
not have been projected at the time of conflict. A hypo-
thesis that has been both violated and supported or one
that has been neither violated nor supported up to the time
in question could not have been projected, since it fails
to satisfy the requirements for projection. In this re-
’gard, the evidence for hypotheses in these two circumstances
is said to be neutral. But if the evidence is neutral for
a hypothesis, then it is also neutral for the contrary of
that hypothesis. It is for this reason that some conflicts
can be ruled out.

It is necessary that the second rule be restricted
to hypotheses with the same antecedents and conflicting

consequences. Otherwise, the rule is too powerful and

48
would eliminate almost all projections of weakly en-
trenched predicates.

Rule 2: "A projected hypothesis with an ill-entrenched
consequent is to be rejected if it conflicts with another
hypothesis (1) that has the same anteCedent and a much
better entrenched consequent, and (2) that is either
(a) both violated and supported or (b) neither."1c3

This rule, however, does not handle difficulties
resulting from conflicting antecedents. Though we have
taken care of projections of ill-entrenched predicates
over the extensions of well-entrenched ones, we have yet
to eliminate projections of well-entrenched predicates
over the extensions of ill-entrenched ones.

Suppose we have a hypothesis, Hi:

All emerubies are green,
where "emeruby" applies to emeralds examined for color
before time t and to rubies not examined before I. Ob-
viously, this hypothesis conflicts with the hypothesis, 2;:
All rubies are red,

and can be eliminated by Rule 1 so long as the latter hy-
pothesis is supported, unviolated, unexhausted, and, thus,
could have been projected at 3. But if we suppose that
fig is either violated or unsupported at the time in ques-
tion, then Hg could not have been projected and could not
have served to eliminate El. Since the antecedents are

not the same, neither could the second rule have elimi-

nated El.

49
A new rule is required to eliminate El on the grounds
of its conflict with the hypothesis, 5:
All emeralds are green.
El and E project the same predicate over different ex-
tensions, since E projects "green" over all emeralds,
while El projects "green" over some emeralds and some

though the positive instances supporting El

rubies. Yet

are the same as thosesupporting E El projects the predi—

9

cate "green" over a larger and less well entrenched ex—
0

tension than 5. Thus, for Rule,3, according to Goodman:
The underlying principle is that where a consequent
could have been projected over the extension of
a given antecedent by a hypothesis, any other hy-
pothesis is illegitimate if it has nothing additional
in its evidence class and yet uses a much less

well entrenched antecedent to_8roject the same
consequent over other things.1

Rule 3: A projected hypothesis is eliminated "when some
other hypothesis with the same consequent could have been
projected, and the antecedent A of the original hypothesis
'disagrees' with the much better entrenched antecedent A'
of the other hypothesis in the following way: although
among things to which the common consequent has been
determined to apply, A applies only to those that A'

applies to, nevertheless A applies to some other things

40
that A' does not apply to."

On the basis of our expanded class of reference

 

hypotheses and with the guide of entrenchment, these three
rules eliminate many of the objectionable hypotheses from

the class of actually projected hypotheses. The remaining

50
members of this class, the new class making up the so-called
"working basis", are all and only those actually projected
hypotheses not eliminated by the application of these rules.
Before hnxfing to the problem of expansion of this working
basis and to other means of eliminating additional objection-
able hypotheses, it should be worthwhile to explain Good- I
man's notion of a "parent" predicate. It will be seen that
parent predicates serve to increase the effectiveness of
the rules of elimination.

As Goodman explains:
...a predicate "P" is a parent of a given predicate
"Q" if "P" applies only to mutually disjoint

classes, and one of these classes is the extension
Of HQ" ,

Thus, "senate" would be a parent predicate of "senator in

the U.S. Congress". One of the mutually disjoint senates
of the world is made up of the senators of the U.S.
Congress.

Increased effectiveness of the rules is brought about
by entrenchment being a hereditary character. Entrench-
ment of a parent predicate is inherited by its descendants.
If we cared to, we might define "siblings" as "coextensive
predicates", since either earned or inherited entrenchment
of a predicate would apply also to any predicates coexten-
sive with it.

At any rate, we can modify and increase the precision

of our methods of gauging entrenchment. If two conflicting

predicates are of about equal earned entrenchment, we shall

51
favor one with a much better entrenched parent predicate.
Furthermore, as Goodman puts it, "the requirement, in
our rules of elimination, that a given predicate be much
better entrenched than another is met if the former either
has much greater earned entrenchment than the latter or
has about equal earned and much greater inherited entrench-
ment."22

We turn now to the problem of expansion of the working
basis to include projectible hypotheses that have not been
actually projected.

Presumptive Projectibility: Expansion

The definition of presumptively projectible hypo-
theses allows the necessary expansion to include all,
though as yet not only, what we want to call projectible
hypotheses. We have available as tools for constructing
the definition the working basis of actually projected
hypotheses, the primitive predicate and rules of elimina-
tion, and, lastly, the innocousZ3"could have been projected".

It should be recalled from our previous remarks on
the general problem of dispositions and our remarks
characterizing the problem of projectibility as Goodman
treats it that the problem of defining the dispositional
predicate "projectible" is one of finding an auxiliary
manifest predicate related in a suitable manner to the
initial manifest predicate "projected" such that this

suxiliary predicate can be used as definiens for "pro-

52
jectible". It will be quite important to notice whether
"presumptively projectible" satisfies these requirements.

The obvious first step in the expansion is that we
want to consider of the unprojected hypotheses only such
as could have been projected, since we can safely ignore
unsupported, violated or exhausted hypotheses. Yet, not
all of this batch of unprojected hypotheses are projectible.
Some we shall want to eliminate through application of
our three rules. The next step, then, is to render the
rules applicable to these unprojected as well as to pro-
jected hypotheses. According to Goodman, this can be
accomplished simply by changing the rules from "a project-
ed hypothesis is to be eliminated (from the working basis)
if..." to read "a hypothesis that could have been pro-
jected is to be eliminated (from the realm of hypotheses
that could have belonged to the working basis) if...".

This eliminates the unprojectible of the unprojected
hypothes:s.

We are now in a position to offer a definition of
"presumptively projectible", which Goodman does as follows:

...a hypothesis could have belonged to the working
basis if and only if that hypothesis is supported
and unviolated and unexhausted and does not run
afoul of disagreements that call for its elimination

by our rules. All and only such hypotheses are
presumptively projectible.

 

Presumptively projectible hypotheses are thus members
of a class containing both projected and unprojected hypo-

theses from which certain objectionable hypotheses have

53
been eliminated. We now have a well-expurgated working
basis for discriminating projectible hypotheses. Yet, it
would be difficult to assess the adequacy of "presumptive-
ly projectible" in a definition of "projectible". First,
our rules will allow us to speak only of comparative
entrenchment. Given any hypothesis, we should be able
by this means to determine whether it is presumptively
projectible, but no clear sense has been established for
the meaning of "degree of projectibility". Second, it
is not the case that we have expurgated our working basis
of all objectionable hypotheses.

Therefore, there is still much to be done, and we
turn now to the question of degree of projectibility
for the resolution of both of these issues.

Degree of Projectibility

 

The difficulty that we now face is that of finding
precise rules to reflect the rather delicate variation of
projectibility between similar hypotheses brought about
by different circumstances. For inmance, the hypothesis

"All marbles in bag B are red"
may be much more projectible than the quite similar hypothesis
"All marbles in bag C are blue"
under different circumstances, though both are presumptively
projectible and contain equally well-entrenched predicates.
To uncover the method Goodman adduces to reflect this

difference, it is necessary that we examine his treatment

54
of initial projectibility index as modified by overhypo—
theses.

The initial index of degree of projectibility is deter-
mined for presumptively projectible hypotheses by compar-
ing the entrenchment of their predicates. Here, of course,
we include both earned and inherited entrenchment. As
Goodman puts it:

If the antecedent or consequent of one such
hypothesis is much better entrenched than the
corresponding predicate of a second, and if the
remaining predicate of the first hypothesis is
no less well entrenched than the corresponding
predicate of the second, then the first has the
higher initial projectibility index.25

This initial index is modified, either positively or
negatively, by the affect of overhypotheses, a method
of bringing to bear what might be viewed as indirect
evidence on the projectibility of hypotheses. Through
overhypotheses we can gauge the affect of certain evidence
not directly supporting or violating a given hypotheses.

Let us illustrate with a certain bagful of marbles B
in stack S. We might have as a presumptively projectible
hypothesis, g:

"All the marbles in bag B are red"
for which we determine the initial projectibility index.
We have examined only some of the marbles in bag B, let
us suppose, but through examination of other bagfuls in
stack S have determined only positive instances of the

hypothesis, E:

55
"Every bagful in stack S is uniform in color."
This hypothesis has, of course, as one ot its, as yet
undetermined, instances, Gl;
"Bagful B is uniform in color,"
which greatly increases the projectibility of 3, so
long as g itself is highly projectible. Again, if g.is
highly projectible, the acceptability of predictions of
its undetermined instances is increased. One of these
undetermined instances is the statement that B is
uniform in color. This, then, increases the projectibility
of any presumptively projectible hypothesis that we have
to the effect that the marbles in B are some one par-
ticular color or other, and as long as g is presumptively
projectible any hypothesis stating that the marbles in
bag B are some color other than red will be violated and,
hence, not pesumptively projectible.
Goodman characterizes such hypotheses as follows:
...a hypothesis is a positive overhypothesis of
a second, if the antecedent and consequent of the
first are parent predicates of, respectively
the antecedent and consequent of the second.26
However, not all overhypotheses have a positive
affect. In some cases, an overhypothesis may serve to
reduce the projectibility of a hypothesis. For instance,
if a highly projectible g were:
"Every bagful in S is mixed in color,"
it would have the effect of reducing the projectibility

of E; In this case, the consequent of g is not the parent

56
of the consequent of a but is the complement of the parent,
since "mixed in color" applies to just those bagfuls of
marbles to which "uniform in color" does not apply. Thus,
the greater the projectibility of this new g, the less
the projectibility of E:

There are, however, conditions on the effectiveness
of an overhypothesis, either positive or negative. First,
it is noticeable in the above examples that the over-
hypotheses are required to have a high degree of projecti-
bility, as for any othergresumptively projectible hypotheses,
is provided by the initial projectibility index as modified
by the affect of overhypotheses. But we are not forced to
retreat through a hierarchy of overhypotheses. Goodman“

points out that g may have an overhypothesis such as E:

"Every stack of marbles in Utah is homogeneous
in color variegation"

with a high decree of projectibility and well entrenched
predicates, but it is not likely that the parents of these
predicates would be well entrenched. Therefore, E's
degree of projectibility would be in terms of its initial
index, unmodified by overhypotheses, since the overhypo-
theses of 2 would not be presumptively projectible.

Furthermore quite similar reasoning can be employed

to show that a hypothesis with a low initial projectibility
index will have a low final degree of projectibility. A

hypothesis with a low initial index can have little of

either earned or inherited entrenchment for its predicates.

57

Yet to say that it has little inherited entrench-

ment is to say that its parents have little earned or
inherited entrenchment. Since any of its overhypotheses

much contain one or another of these ill entrenched parent
predicates, none of the overhypotheses would be presumptively
projectible and could not, therefore, be used to increase

the low initial projectibility index of the hypothesis in
question.

The second requirement is that the overhypothesis be
well supported. This requirement is different from the
first. An overhypothesis can be presumptively projectible,
hence merely supported, without being well supported. But
it does not seem much argumentis required to make this
second point. Obviously, the better an overhypothesis
is supported, the greater degree of credibility is trans-
mitted to its undetermined instances and, thereby, to the
hypothesis of which its predicates are parents.

The third condition is a bit more difficult to present
adequately. Let us follow Goodman's procedure and illus-
trate with an example used previously. Recall that the
hypothesis,fl:

"All the marbles in bag B are red"
was increased in degree of projectibility through the
affect of the overhypothesis, 9:
"Every bagful of marbles in stack S is uniform in color,"

and the support that we had for this overhypothesis from

58
examination of bagfuls of marbles in stack S. But suppose
we attempted to use this support to increase the pro-
jectibility of the hypothesis, El:

"All the marbles in bag W are red,"
where W is in some stack other than S but also in Utah.
We could use our support for the overhypothesis, El:

"Every bagful of marbles in Utah is uniform in color."
But the effectiveness of El in increasing the projectibility
of El is obviously going to be less than the effectiveness
of g in raising the projectibility of a. Therefore, as
Goodman formulates this condition:

Where two equally projectible overhypotheses
bring equal evidence to bear, theqmore specific
one has the more powerful effect.“
The more specific of two overhypotheses is one with
fewer members in the extension of its actecedent predicate,
which is, of course, a parent predicate of the the hypo—
thesis of which the one in question is an overhypothesis.
The importance of these conditions justifies our
being a little repetitious, and I think it is worthwhile
to present Goodman's summary of them:
...the comparative effectiveness of different
presumptively and appreciably projectible
overhypotheses depends upon three factors.
Where such hypotheses are equmly specific and have
equal supporting evidence, their effectiveness
varies with their degree of projectibility.
Where specificity and also degree of projectibility
are equal,effectiveness varies with amount of
support. And where degree of projectibility and

also amount of support are equal, effectiveness
varies with specificity.28

Summary

It will be recalled that in the example taken from
Scheffler the problem was to indicate that "is playable"
may be defined through the specification of such manifest
predicates as "is a record of music or speech", "is of
such and such form", "is of such and such shape", and
"has such and such a history". What Goodman has accomplish-
ed is to specify a set of manifest predicates in terms of
which "projectible" is defined. This set includes the
predicates "unviolated", "supported", "unexhausted",
"could have belonged to the working basis". Any hypo-
thesis to which all of these predicates are applicable
is projectible.

Yet we have not shown the theory of projection to

be applicable to Lewis' problem of the credibility of
phenomenal statements. In the next chapter we hope to
show that an extension of Goodman's theory should readily
solve Lewis' problem.

It is also hoped that in attempting this extension
we shall provide an answer to a criticism of Goodman's
theory suggested by Scheffler. He comments:

The most natural objection to Goodman's new
approach is that it provides no explanation of
entrenchment itself. In using this notion

to explain induction, however, Goodman does not at
all rule out a further explanation of why certain

predicates as a matter of fact become entrenched
while others do not. His purpose is to formulate

60
clear criteria, in terms of available information,
that will single out those generalizations in
accordance with which we make predictions. The
strong point of his treatment is that his criteria
do indeed seem effective in dealing with the
numerous cases he considers.29

If we can show that Goodman has suggested an explanation
of entrenchment that is quite advantageous in itself,
we will have indicated some strength in Goodman's theory

that has perhaps been overlooked.

61
NOTES

1. Goodman (4), Ch. III, pp. 63—83, Ch. IV, pp. 87-120.

. Ibid. p. 65.

3. Ibid., p. 67.
4. Mill (11), pp. 185-6.
. Goodman, Op. Cit., p. 69.

6. For an exposition, in detail, of the difficulties
of the generalization criteria, see Scheffler (l7).

7. Goodman, Op. Cit., pp. 73-4,

8. Note the resemblance of this problem to Lewis'
problem discussed in Ch. II.

9. Goodman, Op. Cit., p. 82.

10. Ibid., pp. 48-9.
ll. Scheffler (18).
12. Ibid., p. 16.

13. Goodman, Op. Cit., pp. 89-90. In terms of Goodman's
figure, the ultimate task of this thesis can now be
described as an attempt to explain this "motion of
the mind" as its tendency to respond to events as sig—
nals of other events. Of the many signaling capacities
objects or events have for the mind some are gradually
sorted out as genuine while others are discarded. We
should like to explain the distinction the mind makes
between valid and invalid projections on the basis of
the distinction between genuine and non-genuine signals.
But first, we shall need to continue with a fairly
detailed examination of the theory of projection. This
may enhance our prospects for a theory of genuine
signality.
14 Ibid., p. 116.
14a. The strategy for the definition of the dispositional
term 'projectible' is, in short, essentially the same
as that employed for the dispositional 'playable'
described on pp. 40-1.

15. Ibid., p. 96.

16. Ibid., p. 96. For an elaboration and extension of the

l7.

thesis of entrenchment as a key to the problem

of induction, to value considerations arising

in classification, see Rudner (14), Part III.

For Goodman's analysis of the locution "could have
been" and why it is required that the hypothesis
be "actual", ee (4), Ch. II, "The Passing of the
Possible". .

Ibid.,pp. 101-2.

Ibid., p. 104.

Ibid., p. 104.

Ibid., p. 105.

Ibid., p. 106.

Reasons "Could have been projected" can be considered
innocuous are presented in (4), Ch. II.

Ibid., p. 108.
Ibid., p.108.

Ibid,, p. 110.
Ibid., p. 115.
Ibid., p. 115.

Scheffler (17), p. 10.

63
CHAPTER V

In this chapter we hope to indicate how the solution
we have been suggesting throughout this thesis may be
given to the problems we have been cOnsidering. Our sug—
gestion is that a predicate of signals may prove adequate
to provide a basis for systematization of some semantical
and pragmatic predicates. We have further suggested that
such an analysis will serve to relate the problems of
semantics and epistemology with which we have been mainly
concerned.

I

Let us briefly elaborate the suggestion we are making.
We noticed that Tarski's definition of truth is not adequate
to solve the problem posed by Lewis on the credibility of
phenomenal statements. We are proposing that an extension
of Goodman's theory of projectibility, supplemented by
Tarski's definition may be adequate for solution of this
problem. Furthermore, we hope to show that the requisite
extension of the theory of projectibility may be accomplish-
ed on the basis of the predicate "is a genuine signal".

It should be noted at the outset that the actual
construction of the suggested system is obviously not goino
to be within our grasp here. What we will do instead is,
first, offer some quite tentative definitions (in the hope

of improving these at some later time) and, second, try

64
to discover whether or not there is any good reason for
assuming that adequate definitions along the lines
indicated cannot be given.

Let us begin by reviewing Goodman's comments in
"sense and Certainty". It will be recalled that the truth
of a statement 2 in the present tense saying "There is now
a red patch in the visual field" is defined in terms of
the truth of an earlier statement F signaling the event
2 describes. B does not signal the event, since signaling
is forecasting; yet, 2 is true if an only if F is a genuine
signal of what P describes. Moreover, and this is the
crucial epistemological point, P is credible ll we are
warranted lg taking F‘lg‘bg a genuine signal: i.e. only
if E belongs to the class of presumptively genuine signals.
Furthermore, as Goodman says of other-tensed cognates of P,
"statements in the past tense can be handled in the same
way as those in the present tense; and tenseless statements,
depending on whether they occur before, during, or after
the time they pertain to, in the same way as statements
in, respectively, the future, present, and past tense."1

First, it is said a statement in the past tense can
be handled in the same way as one in the present tense.

We may thus follow the schema outlined above. Suppose we

have the statement, Pl, that at some time 3 "There was

a red patch in the visual field." Pl is true if and only

65
if El, predicting the occurrence of a red patch in the
visual field at time 3 and occurring prior to time 3, is
a genuine signal.

Second, we have the matter of tenseless statements.
Depending on whether they occur before, during, or after
the time they pertain to, these are to be handled in the
manner, respectively, as future, present, or past tense
statements. .Suppose we have a statement, 33: "A red patch
occurs in the visual field at time 3." If this statement
occurs prior to the time to which it pertains, it is to be
handled in the manner of future tense statements. Thus,
fig is true if and only if it is a genuine signal of the
red patch at 3. If 33 occurs during or after time 3, it
is to be handled in the manner, respectively, of present
or past statements. In both these latter cases, 33 is

true if and only if E} predicting the occurrence of a red

.9

patch at time 3 and occurring prior to time 1, is a genuine

signal. All of these truth conditions are corollaries of

 

jointly the definition of "genuine signal" to be discussed

below and the Tarski treatment of "truth".

 

We have a means, therefore, for relating the truth,
of phenomenal statements in any tense to signaling sit-
uations. Yet, we have two crucial and difficult problems
to consider before we can decide the success of Goodman's

proposal or its extensions. First, we feel that the predicate

66

"is a genuine signal" must be explicated, should be
reduced at least to the broader basis provided by the
predicate "is a signal"; since what we wish to explain lg
(at least in part) "genuine signality". Second, we must
provide means for dealing with statements in the present '
or past tense such that no correlative statement in the
future tense has happened to occur.

These two problems are much like the twofold problem

encountered by Goodman in Fact, Fiction and Forecast. That

 

this is so is indeed the funduental pivot of this thesis.

In Fact, Fiction and Forecast it will be recalled, the

 

 

twofold problem was 33 eliminate from the basis unprojectible

projected hypotheses and £2 expand the basis 32 include

 

unprojected projectible hypotheses.2 We, on the other hand,

 

have the twofold problem of (l) eliminating non:genuine

signals from the basis 2i actual signals and (2) expanding

 

 

 

this basis lg include genuine signals that, though actual,

were not actually signals on appropriate occasions required

 

for the definition 2i truth 9£.9£ credibility; This second

 

 

problem_can be explained as the problem of interpreting a
statement like "F" could have been a genuine signal, if it
had occurred prior to time 1."

In brief, our problem is one of projecting "is a signal"

 

l3 order lg define "genuine signality". Let us attempt

further to determine the relevance of this problem to the

67
theory of projectibility.
II
In Goodman's theory of projectibility, the extra-
systematic explanation of the primitive "is a much better
entrenched predicate than" is given in terms of actual
projections.3 We should recall that a hypothesis is
construed by Goodman to have been actually projected when it
has been used to predict the outcome of examining previously

unexamined cases at a time when it is supported unviolated

9

 

and unexhausted. Some of the undetermined cases of a hypo-

 

thesis may be past cases. Projection of a hypothesis to
such a case is not a prediction. Still, for such a case,
the hypothesis predicts the outcome of examination of
its instance, i.e. the statement describing the past
case. In the sense that outcomes rather than past cases

are predicted we may say that projecting hypotheses is

forecasting.
We might say then that if a hypothesis is actually=

projected, then it is a successful signal of that which it

 

successfully predicts. But we cannot say that if a hypo-
thesisis a successful signal of that which it predicts, then
it is actually projected, since a hypothesis may have been
violated prior to the signaling in question. And a hypo-
thesis that has been violated in one instance may still be

used to predict the outcome of examining some other instance.

68
To illustrate, suppose a hypothesis, El, says:
"All the marbles in bag E are blue."

Of the many marbles from bag E so far examined, all but
one have been found to be blue. Still, El may success-
fully signal the outcome of examining another marble
from bag E. Thus, El is a successful signal, though a
replica of El failed to signal the outcome of examining
the non-blue marble.

The sense of "replica" intended at this point is that

/

employed by Goodman in The Structure of Appearance.“1 In

 

this sense, gEly linguistic events are replicas of them-
selves and some other linguistic events, e.g. any occur-
rence of "Paris" is a replica of itself and also of any
other occurrence of "Paris".

We might avoid the above difficulty by refusing to
acknowledge as a signal an event a replica of which has
failed on some occasion to be a successful signal. This,
however, would be a costly choice, since it would serious-
ly restrict the breadth of our basis.

In order for our system to have important relevance
to epistemological problems, we must start with events ac-
tually :taken to be signals of other events. We want
accordingly to use "E is a signalof E" in such a way as
"E is taken to be a signal of E and E appropriately occurs".
We do not want to use it in such a way that we are commit-
ted to every replica of E signaling every replica of E.
The step from "E is a signal" to "E is a reliable signal"

is part of the way from "A is a signal" to "A is a genuine

69
signal". It is just this step we should like to explain.

Yet, it does seem we are on the right track to deter-
mine the reliability of a hypothesis as a signal on the
basis of the success of its replicas. Following this
procedure may also suggest a way of distinguishing non-
genuine signals in the basis of actual signals.

We should like to define the predicate "is more
reliable as a signal than" in such a way that the minimum
qualifications for a hypothesis to be a reliable signal
is that it be supported, unviolated and unexhausted or
that no instances of a reliable hypothesis have failed
successfully to signal that which was predicted. Our con-
tention will be that projectible hypotheses and genuine
signals are at least reliable signals.

As a definition of successful signaling we propose:

(i) For all replicas of an instance, E, of a hypothe-

sis, H: replicas of A are successful signals of

repliEas of E if and Bnly if some replicas of E
are signals of replicas of B, and any previous

replica of E that is descrihed by an.&nstance of
E could have signaled a replica of E.8

The locution "could have signaled" is used here to mean
that if E and E are any replicas of E and E, respectively,

either 9 did signal E or 9 occurred just prior to E and

 

lacked only actually being taken E§.§ signal gl E, and

 

there have occurred some instances of signal taking.5
This condition serves, then, at a stroke to effect

both a requisite expansion and a requisite contraction of

 

the basis. Employing it, we may eliminate any signal that

70
has replicas which on occasions have been unsuccessful
as signals. On the other hand, it expands the basis to
include some nonésignaling events that could have been
signals.

It should be noted that what we are attempting to do
in this definition is to match Goodman's requirement that
an actually projected hypothesis be both supported and
unviolated. We have done fairly well in a matching of the
requirement that a hypothesis be supported, but we have
somewhat modified the requirement that it be unviolated.
On our view the only hypotheses rejected as unsuccessful
signals are those that have been explicitly used to predict
the outcome of examining previously unexamined cases and
have failed to do so. We have not taken account of infor-
mation we might have of negative instances when no such
prediction was actually made, nor of the more difficult
Case of signalings we should, perhaps, like to call suc-
cessful though the signaled event does not occur as signal-
ed. However, to take account of these problems would
seriously complicate this exposition and not greatly help
us accomplish our more modest ends. We are not, after all,
pretending to offer final, adequate definitions. We SEE
attempting to suggest a program for at least some initial
steps in giving such definitions.

With these qualifications, we are now in a position

to suggest a definition for reliability.6

 

7l
At some time 1:

(ii) A is more reliable as a signal than C if and only
if:
(1) the number of replicas of A that are signals
of replicas of that of which rep.licas of A are
successful signals, is greater than the number
of such replicas of C;
(2) both replicas of A and replicas of_ are
successful signals; and
(3) A is taken to be a signal of a replica of
that of which replicas of E are successful sig-
nals.

We may drop the requirement that both replicas of E
and replicas of E be successful signals, if we decide that
at this point we are discussing only successful signals.

We are attempting to say in this rather complicated
fashion that the reliability of a signal is determined
on the basis of the success of its replicas. The final
requirement in the definition, that E'be taken to be a
signal, is inserted for two reasons. First, it would
seem odd to say that E is more reliable as a.signal than
9, when E is not taken to be a signal at all. Second,
this requirement parallels Goodman's condition that a

hypothesis be unexhausted at time of actual projection.

However, our condition is weaker in that it is not re-

 

quired that a hypothesis BE unexhausted, only that it be
£3532 to be unexhausted at the time in question.

However, it is likely that the manner in which we
have so far defined "successful signal" and "reliable as
a signal" will nonetheless determine a class of actually
projected hypotheses with fewer members than the class of

actually projected hypotheses Goodman uses as his basis.

72
This is likely because our definitions require that

replicas both of an instance of the hypothesis and that

 

which was predicted occur in order for the hypothesis

 

to be a successful signal. To illustrate, suppose El,

"All the marbles in bag E are blue,"
is explicitly adopted and successfully predicts the out-
come of examining the next marble drawn from bag E. On
our view, this amounts to a case of signaling only if a
replica of both

"The next marble drawn from bag E will be.blue"
and i

"This marble is blue"
(or some other phrasing of an instance of El and a report
that a given marble is blue) occur at appropriate times.
In circumstances such that one or the other of these fail
to occur, El has not been an actual signal on our view,
though for Goodman it has been projected in these cir-
cumstances.7

The remark that the reliability of a signal is deter-

mined on the basis of actual successes is similar to Good-
man's explanation of entrenchment. He emphasizes, it will
be recalled, that entrenchment of predicates is determined
on the basis of frequency of actual projection. This

similarity suggests deepening Goodman's study by attempting

 

 

 

to define entrenchment gfi predicates lg terms of reliabi-

lityugg signals.

According to our proposed definition of "reliable",

73

replicas of El are more reliable than replicas of E2
if and only if El has been more frequently used for
predicting than EE. We say, therefore, that El is more
reliable than EE if and only if its replicas are more
'reliable than those of EE. Thus, up to time l:

H1 has been more frequently projected than EE

if and only if El is more reliable as a

signal than EE.

There is no parallel to this definition in Goodman's
theory. It is required here in order to iefine the
frequency of projection of predicates, which Goodman uses
to explain earned entrenchment.

Earned entrenchment in Goodman's theory is entrench-
ment of a predicate ii an antecedent or entrenchment of
a predicate EE a consequent in a general conditional.
When we speak of "antecedent predicate E", we mean "the
predicate £.E§ an antecedent or component of an antece-
dent". "Consequent predicate E" means "the predicate E
ii a consequent or component of a consequent".

Up to time l:

(iii) Antecedent predicate E has been more frequently

projected than antecedent predicate if
and only if hypotheses with antecedent pred-
icate E have been more frequently projected
than hypotheses with antecedent predicate O.
Also, up to time l:
(iv) Consequent predicate E has been more fre-
quently projected than consequent predicate
Q if and only if hypotheses with consequent
predicate P have been more frequently pro-

jected thafi hypotheses with consequent predi-
cate Q.

74
Thus, the reliability of a hypothesis attaches to
the predicates of which it is composed. We may now propose
a definition of "entrenchment".
To save space, the following is to be understood as

applying mutatis mutandis to the comparison of antecedent

 

predicates to antecedent predicates or consequent predi-
cates to :consequent predicates.
At time 1:

(v) Predicate E is better entrenched than predi—
cate E if and only rif predicate E has been
more requently projected.than predicate 3.
After introducing his primitive relation (see p.
45 above), Goodman makes the following comments:

Like Hume, we are appealing here to past re-
currences, but to recurrences in the explicit
use of terms as well as to recurrent features of
what is observed. Somewhat like Kant, we are
saying that inductive validity depends not only
upon what is presented but also upon how it is
organized; but the organization we point to is
effected by the use of language and is not at-
tributed to anything inevitable or immutable
in the nature of human cognition. To speak very
loosely, I might say that in answer to the ques-
tion what distinguishes those recurrent features
of experience that underlie valid projéctions
from those that do not, I am suggesting that the
former are those features for which he have adopt-
ed predicates that we have habitually projected.8

Like Goodman, we have been concerned so far with the
predictive use of language and have framed our definitions

with the view to define his predicates. Yet the first

,

sentence of the quoted paragraph suggests a way of greatly

fortifying our analysis. The signal relation holds among

 

non-linguistic events as well as among linguistic and

 

 

75

non-linguistic events. We can expand our analysis by

 

taking into account non—linguistic signals. This step
is required if we want to preserve the breadth of our
basis.

We could modify the definition of entrenchment
to read "predicate E and predicates coextensive with it"
and "predicate g and predicates coextensive with it" in
place of "predicate g" and "predicate g". Then, we could
proceed to follow Goodman's analysis in order to define

"genuine signality" in terms of "projectible". But this

_._....._____.£2

 

course is prevented by our goal of definincT "projectible"
in terms of "genuine signality". -
III

Though we have indicated how a bridge might be made
between a predicate of signals and Goodman's theory of
projectibility, instead of attempting to follow through
‘in the manner suggested above, we turn now to the task
of admitting non-linguistic signals by modifying our

definitions to admit non-linguistic as well as linguistic

 

signals.

In order to accomplish this, we shall extend the sense
of "replica" employed here, to cover “non-linguistic events.
Each event of the events that constitute the extension of
a predicate is, in a sense, a replica relative to that
extension, of itself and of any other event in the exten-

sion of that predicate. We shall, in fact, say that events

76
are replicas of one another relative to some actual
predicate that applies tx;aoth. Now, the definition of
"successful signal" may be modified so that ceunted as
successful signals may be both non-linguistic members
of extensions of predicates as replicas of each other 32d
linguistic replicas.

Correspondingly, we shall speak elliptically of 3
being a successful signal of g, and construe this as
meaning that members of the extension of R are success-
ful signals of members of the extension of g, and "P-
replica" means "a replica in the extension of g of a member
of the extension of E".

Supplementing the definition already offered we
have:

(vi) E is a successful signal of . if and only if

some E—replicas are signals of g-replicas,

and any knowg g-replica could have signaled

a g-replica.
On this basis we may attempt to define a sense of "rel-
iable" other than that previously considered. Suppose
we have a hypothesis, fl:

"Whenever 2 occurs, then g occurs.

It has been suggested this be interpreted as stating that
members of the extension of E are successful signals of
members of the extension of g.

It may occur, for instance, at some time of occur-
rence of a B-replica, that E signals that a B-replica

is a signal of a g—replica. In this case, fl is a reliable

77
signal if and only if g is a successful signal of
Q. And, in general, we may say:

(vii) Any signal that one event is a signal of another
event is reliable if and only if replicas
of the former are successful signals of
replicas of the latter.

III

We have examined the question of the relevance of

 

our problem, of explicating "genuine signality", to Good-
man's theory of projectibility. Let us now turn to an
investigation of our problem.

Our new predicate "is a reliable signal" has an ex-
tension of "observed regularities", including some "regular—
ities" in the predictive use of language, which is similar
to Hume's basis. Taking this predicate, together withm
those previously defined, we may now attempt to solve the
twofold proolem with which we began this chapter. It
was remarked that this problem is analogous to Goodman's

twofold problem in Fact, Fiction and Forecast, and we

 

shall, therefore, follow his lead in order to propose
a solution to the problem we are facing.

As previously mentioned, our twofold problem is that
of eliminating non-genuine signals from the working basis
and expanding this basis to include genuine signals that,

though actual, were not actually signals on appropriate

 

occasions. We have as our working basis all signals that
are reliable according to (vii). In order to avoid con—

fusion, let us differentiate the two senses of "reliable"

78
so far defined by calling (ii) "reliablel" and (vii)
"reliablez". I

It will be a complicated task (and, moreover, one
for the psychologist) to specify the numerous psychological
factors involved in signal-takings.' We shall mention
some of these while trying to avoid speculating on the
results of future psychological research. An event oc-
curs and we note it. We have to choose whether as a sig-
nal totake it as a E-replica, g-replica, etc.. This
choice will be influenced by any number of considerations%1
but primarily, perhaps, by (l) to what further events of
interest we could take the occurring event to be, say,
instrumental and (2) how reliable it is as a signal of
those interesting events to which we take it to be instru-
mental. Such decisions require, most often, a rather
delicate and complex balancing of the information avail-
able. We should like to investigate some of the sort
of information that is available for the second of these
considerations.

Let us say an event 5 occurs that is primarily noted
as a E-replica and an B-replica. (We take, for economy of
exposition, all other predicates applicable to this event
to be “uninteresting" in our value considerations.)

First, let us say, we scrutinize whether A as a g-
replica is a reliablel signal of a g-replica and whether

A- as an gereplica is a reliablel signal of an §~replica.

79

Rules are required in order to enable us to eliminate
unreliable signals. And, since we intend at a later
point to equate "projectibility" and "genuine signality",
we shall follow the rules Goodman invokes to eliminate
unprojectible hypotheses from his basis.
Bple 1: g as a signal that a B-replica is a signal of
a Q-replica is to be ruled out as a reliablez signal if:
(l) i as a signal that an g-replica is a signal of an §-
replica is more reliablel than 5 as a signal that a B-
replica is a signal of a g-replica; and
(2) it is not the case that both g is a reliablez signal
that a B-replica is a signal of a Q-replica ggd_é is a
reliablez signal that an g-replica is a signal of an
S-replica. .

This rule may be taken to correspond roughly to
Goodman's rule la, (p. 47 above), though our rule 1 is

more explicitly applicable to the extensions of predicates

 

than is Goodman's rule la.

Rule 2: g as a signal that a E-replica is a signal of a
Q-replica is to be ruled out as reliablez if:

(1) A is also a signal that a E-replica is a signal of a
I-replica;

(2) the number of replicas of 5 that are signals of :-
replicas is greater than the number of replicas of a that
are signals of E-replicas;

(3) either (a) there is a replica of g that was taken to

80
to signal a I-replica and failed to do so, or (b) neither
has there been a replica of A that was taken to signal
a T—replica and failed to do so, nor has there been more

than very few13

replicas of & that are signals of :-
replicas; and

(4) it is not the case that both A is a reliablez signal

 

that a E-replica is a signal of a Q-replica 32§.£ is a
reliablez signal that a E-replica is a signal of a :-
replica.

This rule is taken to correspond roughly to Good—
man's rule 2 (p. 48 above). >In our schema it would be
difficult to meet both of the requirements that the con-
sequent predicate of the conflicting hypothesis be well-
entrenched, and that the hypothesis be neither violated
nor supported. We, therefore, substitute for the latter
that a signal be neither violated nor wgll-supported.
Perhaps, the degree of neutrality of the available evi-
dence in our formulation is less than in Goodman's, but
the difference in effectiveness between "barely supported”
and "not at all supported" in this case is likely to be
negligible.

Rule 3: é'as a signal that a g-replica is a signal of a
Q-replica is to be ruled out as reliablez if:

(1) fl is a signal that an R-replica is a signal of a g7
replica;

(2) an R-replica is more reliablel as a signal than a

81
B-replica;
(3) any previous g-replica is both an B—replica and a
B-replica;
(4) any previous R-replica is a g-replica;
(5) there are P-replicas that are not R-replicas; and

(6) it is not the case that both A is a reliableg signal

 

that a E-replica is a signal of a Q-replica igd.é is a
reliablez signal that an R-replica is a signal of a
Q-replica.

Now we have as our working basis all and only such
reliablez signals that do not violate one or another of
the above three rules. We also have a great many pro-
blems. The ambiguity we have allowed throughout, i.e.
whether events, or characteristics, or occurrences of
characteristics construed as other than events, are
signals, may ruin our efforts at this point if some care
is not taken. We must also take care that the dual sense
of "replica" does not lead us to take these rules to be
other than they are.

To illustrate these difficulties, we must,recall that
we have not specified whether "A" in the above is an

.event, i.e. an occurrence of a (i.e.,one) characteristic.

 

Also, the dual role of "replica" leaves it in doubt
whether a replica of a is a non-linguistic event, a lin-
guistic event, or either a linguistic or a non-linguistic

event. Obviously, these two problems are not independent.

I

82
If we take A to be an occurrence of just one characteristic,
we thereby limit the variety of its replicas to one,
thus eliminating the last of the alternatives. To avoid
this, and also to offer a means of covering known conflicts,
we take A to be an occurrence of characteristics; and
if A is a linguistic event, its replicdSare linguistic
events, if_A is a non-linguistic event, it has replicas
that are non-linguisitc events 32d replicas that are
linguistic events.

We explain this latter interpretatkan as follows:
we are taking 5 to be an occurrence of a complex of
characteristics, some of which may be "account—takings"
of other characteristics of the complex. To speak of it
loosely, some of these account-takings, when non-linguistic,
may nonetheless be incipient speech or speech patterns.

We should like to hold that the similarities of these
occurrences to linguistic occurrences justifies taking
some of these as replicas of each other. We do not want
to commit ourselves on psychological theories of incipient
speech, yet, taking the above in the way we have seems a
promising course available to us.

On the other hand, if 5 is a complex containing only
linguistic occurrences, either written or uttered, we need
not assume it has non-linguistic replicas.

These considerations enable us to improve upon some

remarks made earlier. It would seem from the way in which

83
these rules were introduced that they are applicable only
to occurrences like "fl" is a signal that g is a signal of

Q. We have offered, as yet, no clarification of "is a

signal that" which is, prima facie, a quite different

 

predicate from "is a signal of". Further, we have not
shown how our rules can be used to determine the reliabilityg
of such occurrences as "B is a signal of Q". We now
assume that any occurrence like "3 is a signal of E" can
be taken to be an occurrence of "g is a signal that g
is a signal of Q". The rationale underlying this assump-
tion is that the difference between the two predicates is
primarily one of emphasis. To see this, it should be
remembered that earlier in this thesis we were concerned
with the occurrence of a characteristic as a signal of
another occurrence. Now, though we are still concerned
with the characteristic, we are interest, primarily, in the
taking of it as a signal. It is hoped these remarks serve
both to clarify the locution "is a signal that" and to admit
as reliableZ signals all occurrences where E is a signal
of 9 that do not run afoul of one or another of the three
rules.
I I V

We have still to eliminate from the working basis all

reliablez signals such that no conflicting signals, as a

matter of fact, happened to occur, yet are such that a

conflict could have occurred. We must also expand our

84
working basis to include genuine signals that could have
occurred. In both these cases, we assume merely that
the conflicting signal and the reliable2 signal are
actual, i.e. occurred at §3m3_time.-

Any signal in the working basis is to be eliminated
from the working basis if there is any actual signal
that conflicts with it according to the rules if the
latter had occurred at the time of the former. The
conflicting signal then is the Egfigof the latter actual
signal and the time of the former signal.

All and only the signals 22: eliminated by our

xpanded rules are presumptively genuine signals. We

 

propose the following as a definition for "is a genuine
signal":
(viii) Any signal is a genuine signal if and only
if it is a presumptively genuine signal,
and that of which it is taken to be a signal
occurs.
We have,finally,the remaining problem of expanding this
basis to include genuine signals that could have occurred.
(ix) If a certain event 3 occurs and an event é
occurs before E, then é could have been a genuine signal
of E if and only if:
(1) é could have been a signal of g (in the sense of
"could have been a signal" previously characterized), and

(2) A does not run into conflicts accordin-~ to our expanded

W
L)

4

rules. If B occurs and A occurs durin: or after B, than

C

A could have been a genuine signal of E if and only if A

85
could have occurred prior to E (the sum of g and some
time 1 prior to g is an individial) and i does not run
into conflicts according to our expanded rules.

V

The nexus of the two sets of problems previously
considered, i.e. those of defining certain semantical
predicates and those of solving certain epistemological
problems, can be found in the definition of "is a
genuine signal".

We have already investigated the dependence of the
problem of analyzing epistemological predicates on a
definition of "projectible". Roughly in our terminology,
if a hypothesis is projectible, then it is a presumptively
genuine signal; and if a linguistic event is a presumptive-
ly genuine signal, then it is a projectible hypothesis.
[There are difficulties in the latter half of this con-
junction, but I think the definition of "linguistic event",
given in such a way that a linguistic event has no non-
linguistic replicas, will handle some of them.]

Also, defining "genuine signal” in such a way that
that which E signaled occurs enables us, as indicated in
Part I of this chapter, to define "true" for phenomenal
statements, though it is not this consideration that
allows a solution to Lewis' problem.

It will be recalled that Lewis was concerned with

86

the here and now credibility of phenomenal statements and

 

construed Goodman as offering a "predictive" criterion for
judging such credibility. However, he mistakenly thought
that Goodman was speaking of the predictiveness Ef phenomenal
statements. An explication of the credibility of phenomenal
statements suggested by Goodman in "sense and Certainty"
may be offered in terms of the sense of "genuine signal"
developed here.
For a phenomenal statement 3, "there is now a red

patch in the visual field", occurring at time i:

P is credible, i.e. P may justifiably be believed

to be true, if and only if red patch at E was
signaled by a presumptively genuine signal.-

 

For the explication of "true? as it appears here, we can
rely upon the Tarski definition and, hence, bring in another
strand of the analysis we have followed.

Thus, we have indicated how an extension may be made
to Goodman's theory of projection. Further, we have explored
how this extended theory, supplemented by Tarski's definition
of "true", may be adequate to resolve Lewis"problem on the
credibility of phenomenal statements. We have done so with
the intent to show that this problem can be handled within
the framework of empiricism and that there is, Lewis to the
contrary notwithstanding, no theoretical requirement within

such a framework for the indubitability of phenomenal state-

 

ments.

Projectible hypotheses lend projectibility to their

87
instances, such instances when signaling phenomena lend
credibility to phenomenal statements. A phenomenal
statement is true if the instance predicting what it
describes is a genuine signal, and a true instance lends
confirmation to a projectible hypothesis of which it is
an instance.

VI

Many of the further problems solution of which is
still required to render the suggested analysis wholly
adequate have “already been exposed. It would not further
our proposals a great deal to elaborate on these problems.
Yet, it seems to the writer that all of the problems that
we have exposed have been such that techniques EEE avail-
able for their solution. I think, in fact, that we can
conclude, with one reservation to be noted, that we have
found no good reason to assume an adequate analysis of
these problems cannot be given.

The reservation that we have is that we have so far
not commented on the problem of an extrasystematic explana-
tion of the primitive predicate "is a signal of". We have
not done so because, in a sense, it is not part of our
philosophical problem. As Goodman notes of the signal
relation, "so long as we are satisfied that the relation
clearly obtains in nonlinguistic experience, we can post-
pone consideration of its anatomy and genealog ."15 A

psychological account of signaling is not part of our task.

88

Yet we should mention what any such account would at

.
least be required to produce in the way of explanation.
In the way of extrasystematic interpretation, the
signaling relation will require at least biological,
physiological, anthropological and psychological theories
to account for the values "things" have historically had
for our species in the course of its development, how we
learn to discriminate among signals, whether we ESE
values i3 things and how we thus learn to modify our
behavior to render these values accessible to us. All
these questions and many more require answer for an
adequate systematic explanation of signaling and expli-
cation of "signaling". Yet, I think we can decide that

any explanation would be inadequate if it fails to ac-

knowledge that someone takes account of something and

 

that that something may be either a linguistic or non-

linguistic event.

10.

11.

89
NOTES
Goodman (5), p. 166.
See Ch. IV above.
See Ch. IV above.
Goodman (6), pp. 287-90.

We shall not be concerned to examine the many faults
of this proposed definition in detail, since it is
ultimately to be replaced. See p. 76 below.

We will avoid becoming entangled in problems on the
maximum or minimum duration of signaling events.
These vary from one species to another and is too
involved a problem for us to treat here. It is
assumed, however, that such a treatment would be
required for an adequate extrasystematic inter—
pretation of the signaling relation. Here, we
assume merely that signaling occurs over a brief
duration.

This definition is intended, as was implicit in (i),
to be temporally qualified.

We shall not discuss the circumstances under which
a replica of either of the above could be uttered
in order to qualify Hl as having been legitimately
a signal in such a cage.

Goodman (4), pp. 96-7.

This definition needs to be improved in at least one
important way. As it stands, it fails to count as
successful signals occurrences of similar character-
istics such that there actually is no predicate ap—
plicable to them and only to them. An adequate anal-
ysis should handle the problem of specifying cir-
cumstances under which such characteristics could
have been the extension of a predicate. We need

not be concerned with this problem here, since we
have already indicated a basis broader than that of
the theory of projection.

Since an adequate definition of "is a reliable signal"
should isolate a working basis similar to that which
Hume employed, we can now more fully understand the
rationale of Goodman's vindication of Hume. For fur-
ther light on this issue, see Rudner (14), Part III,
and Scheffler (17).

A clarification of this situation can be found in
Rudner (l4).

12.

13.

14

15.

90

Throughout this chapter, we have allowed ambiguity

as to whether we are taking signals to be events,

or characteristics, or occurrences of characteris-
tics construed as other than events. It does not

seem we have as yet good reasons for making a choice
among these alternatives. We shall attempt to present
such reasons and resolve the ambiguity on p. 81.

The decision as to how many are "too few" and how

many "enough" involves consideration of the importance
of being mistaken. Rudner makes this point clearly

in (13) and (14), Part I.

The use of "sum" here is intended to be that defined,
for one place, in (6), Ch. II, Part 4, pp. 42-51.

Goodman (5), p. 165.

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

(11)

(12)

(13)

(14)

BIBLIOGRAPHY

A. J. Ayer, "The Criterion of Truth", printed in
Philosophy and Analysisy edited by Margaret Mac—
Donald, Philosophical Library, New York, 1954, pp. 237-41.

 

Gustav Bergmann, Meaning and Existence, The University
of Wisconsin Press, Madison, Wis., 1960.

 

Max Black, "The Semantic Definition of Truth", printed
in Philosophy and Analysis (See (1) above), pp 245-60.

 

Nelson Goodman, Fact, Fiction and Forecast, Harvard
University Press, Cambridge, Mass., 1955.

 

Nelson Goodman, "Sense and Certainty", The Philosophi-f
cal Review, Vol. LXI, Number 2, April, 1952, pp. 160—7.

 

 

Nelson Goodman, The Structure of Appearance, Narvard
University Press, Cambridge, Mass., 1951.

 

C. I. Lewis, An Analysis of Knowledge and Valuation,
The Open Court Publishing Company, La Salle, Ill.,_l946.

 

C. I. Lewis, "The Given Element in Empirical Know-
ledge", The Philosophical Review (same issue as (5)above),
PP. 168—75. ‘

 

C. I. Lewis, Mind and the World-Order, as reprinted by
Dover Publications, Inc., New York, 1956.

 

R. M. Martin, Truth and Denotation, The University of
Chicago Press, Chicago, 111., 1958.

 

J. S. Mill, A System of Logic, as edited by Ernest
Nagel and printed under the title John Stuart Mill's
Philosophy of Scientific Method, Hafner Publishing
Company, New York, I950.

 

 

 

Charles Morris, Signs, Language and Behavior, Prentice-
Hall, Inc., New York, i946.

 

Richard Rudner, "The Scientist Qua Scientist Makes
Value Judgments", Philosophy of Science, Vol. 20,
No. 1, January, 1953.

 

Richard Rudner, "Classification, Induction and Uses
of Value Judgments in Philosophy", paper read at a
meeting of the Fullerton Clue in May, 1955, at Bryn Mawr.

(15)

(16)

(17)

(18)

ii

Richard Rudner, "Sign Process and Valuation: A Reply
to Copi", The Journal of Philosophy, Vol. LV, No. 8
April, 1958.

 

Alfred Tarski, "The Semantic Conception of Truth",

as reprinted in Semantics and the Philosophy of Lang—
uage, edited by Leonard Linsky, The University of
Iiiinois Press at Urbana, 1952.

 

Israel Scheffler, "Inductive Inference: A New Approach,"=
Science, Vol. 127, No. 3291, January, 1958.

Israel Scheffler, "Prospects of a Modest Empiricism",
The Review of Metaphysics, Vol. X, No. 3-4, 1957.

 

“III ifill’iujlflfijll W113 imam“