The paradigm-case argument -- GRICE and URMSON on some questions concerning validity

Speranza

In "Studies in the Way of Words", Grice refers to Urmson's use of the "paradigm-case argument'.

The programme of Grice's and Urmson's project is, more or less, as follows:

A) First of all they outline a method of "argument "which is often used in the work of philosophers, and which they take to be a "perfectly legitimate" type of argument in its proper place and
when used for its proper purpose.

This method of "argument" we may call the appeal to the standard example, or to the "paradigm case".

By the "paradigm-case argument", the philosophical doubt whether something is really an X is exposed as being in some way "improper" or "absurd" by means of a demonstration that the thing in question is a standard -- or as Plato would prefer, "paradeigmatic" -- case by reference to which the expression 'E ' has to be understood, or a doubt whether anything is "E" is exposed by showing that certain things are standard cases of what "E" in question is designed to describe.

B) In the second place, Grice and Urmson show how the "paradigm-case argument" has been
applied by some philosophers to the problems of the validity of deductive, inductive, and ethical argument, here, as later, with special reference to the latter two.

As a first rough approximation one can say that the "paradigm-case argument" has been used to show that at least some arguments in each field must be valid since they are standard examples of validity in their sphere, by reference to which validity in that sphere must be elucidated, so that to query
their validity must be absurd and improper.

C) Thirdly, Grice and Urmson try to show that because of the evaluative element in the meaning of 'valid ' the argument from standard cases cannot be applied to these fields without considerable
modification, and that after such necessary modification the argument does not prove all that many who have used it have wished to prove, though it is not entirely without weight.

In particular, when thus modified the argument leads to a better formulation of the problems which it fails to solve.

This is no mean achievement in such difficult fields.

(A) THE NATURE OF THE "PARADIGM-CASE ARGUMENT"


Suppose that someone looking at a pillar-box expressed a doubt whether it was really red.

He might indeed express doubts whether it was really red because he thought that the light was unusual, or that his eyes were bad, or something of that sort.

But suppose that he expresses doubt for none of these reasons but doubts whether "red" can properly apply to this sort of thing.

We would then be at a loss and probably ask him what (on earth) he MEANT by "red" if he was unwilling to call this red, or say that by *red* we meant being of just some such colour as this ' If we do not call this red then what would we ?'

Thus, using a simple form of the argument from standard examples, we can make him see that
there is something absurd in his question, since there is no better way of showing what the word 'red' means than by pointing to things of this colour.

Now a slightly more sophisticated example to show what is here meant by the argument from standard examples.

In his popular book, The Nature of the Physical World, in an
attempt to bring out in a vivid fashion the difference between
the scientific and everyday description of such things as
desks, Eddington said in effect that desks were not _really_
solid.

Susan Stebbing, in her book Philosophy and the
Physicists, used the argument from standard examples to
show that this way of putting things involved illegitimate
mystification.

This Langer did by simply pointing out that if
one asked what we ordinarily mean by "solid" we immediately
realize that we mean something like 'of the consistency of
such things as desks'.

Thus she showed conclusively that
the novelty of scientific theory does not consist, as had been
unfortunately suggested, in showing the inappropriateness of
ordinary descriptive language.

Two comparatively trivial uses of the argument from
standard examples have been given to illustrate its character.
Obviously to give more subtle examples would require much
space, and at present we are concerned only with what the
argument is, not with showing that it is valuable or important.

But as Grice and Urmson attack what they
consider to be illegitimate uses of the argument it would be
as well to say now that, usually in conjunction with other
techniques if this type of argument can be used to clear up
a number of vexatious philosophical problems.

In par-
ticular, it can be used time and time again to show that
problems have traditionally been incorrectly formulated ;
and every philosopher knows how important correct formulation of problems is.

It is to be hoped that enough has been said to show the
general character of the argument from standard cases. A
full discussion is not possible in the space available.

(B) THE PARADIGM-CASE ARGUMENT AS APPLIED TO "VALID"

Grice and Urmson then set out the use of this argument to solve some
major problems about validity ; in the first instance this must
be done as persuasively as is consistent with extreme brevity.

As Grice and Urmson think that these uses of the argument are, in part, illegitimate, their argumentation will at this stage embody error : but we cannot expose error until we have it before us, pre-
ferably in an attractive guise.

Grice and Urmson first apply the argument to deductive reasoning, since it is instructive to compare
its use here with other uses; we shall then apply it to in-
ductive and ethical reasoning, which are for us now the
central issues.

The reader is asked to bear in mind through-
out this section that I am putting a case, not expressing my
own opinions.

(a) Deductive reasoning

Logicians, in so far as they have been concerned to understand the nature of reasoning and not to produce abstract calculi or * languages *, have not been producing arbitrary fiats for us to obey when they have put forward principles of valid inference.

Nor again, do these principles present themselves as truths independent of actual reasoning.

The logician is attempting to make explicit principles of validity already
implicit in our judgments of the validity or invalidity of actual
arguments.

Thus it is not the case that the validity of the
syllogistic arguments is determined by their conformity to the
rules of the syllogism.

It would be more correct to say that
suggested rules of the syllogism are to be accepted only if
they recognize as valid what we would in any case recognize
as valid, and nothing" else.

Rules of inference and principles
of validity have to be abstracted from standard examples of
valid arguments.

A suggested rule is refuted if it makes
valid a standard example -of an invalid argument, or vice
versa.

This being so, it is simply meaningless to ask whether
the standard examples of valid deduction are really valid,
or to ask whether any deductive arguments are valid.

The
meaning of "valid" with regard to deductive argument is
determined by these standard examples they are the ulti-
mate standard and court of appeal.

To ask whether a tricky
argument is or is not valid is in the end to ask whether it is
like the standard examples of valid arguments or not,
though our direct appeal may be to principles.

(b) Inductive reasoning

Perhaps no one has wished to query the validity of
straightforward deduction, so that our last paragraph may
have seemed to stress the obvious.

But we may argue
similarly with regard to inductive reasoning ; and the general
validity of inductive reasoning has indeed been questioned ;
this is the celebrated problem of induction.

Now sometimes
when the general validity of induction has been questioned
the doubters have indeed had an independent criterion by
to judge it ; they have relied on the principles of valid
deduction.

But this is clearly an error.

For inductive arguments are supposed to be inductive, not deductive, whereas
if they answered to the criteria of deductive reasoning they
would be deductive.

A perfectly good cat would
get low marks at a dog show,
and be none the worse for that.

But
if we abandon the irrelevant criteria of deductive validity,
how are we able to condemn all inductive reasoning as
invalid?

Whence come our principles of judgment ?

Do
we not come near to the evident absurdity of saying that all
men are abnormal, all perception illusory?

When we in fact
regard an inductive argument as invalid it is because it
differs importantly from those which we regard as valid ; a
contrast is intended.

Here then, it appears, as in the case of deductive reasoning, we must start from the standard examples of valid and
invalid argument and elicit from them our principles of in-
ductive validity.

It is not required that the scientific reasoning
of a Newton or a Pasteur should conform to text-book canons
but that text-book canons should be based on a study of them.


Except by reference to such examples no meaning can be
attached to the term ' valid' in the sphere of inductive argu-
ment.

If the validity of such examples is denied, by what
standards is it being judged ? If the irrelevant standards of
deduction have not been dragged in, then surely there are no
standards available.

These examples set the standard.

That is how the argument runs with regard to induction.

The problem whether any inductive arguments are valid is held
up as absurd, and with it the traditional problem of induction.

(To check Grice's and Urmson's exposition of this argument readers are be referred to a few
expositions of it.

It is given in extenso in * Bertrand Russell's Doubts about Induction*, by Paul Edwards, Mind, 1949 (reprinted in Logic and Language >
Vol. I).

For shorter versions see, e.g. the last chapter of An Introduction to Logical Theory, by P. F. Strawson, and A. J. Ayer's introduction to British Empirical Philosophers, pp. 26-7).


(c) Ethical reasoning

It is clear that we can use an exactly similar technique
to prove that it is absurd to doubt generally the validity of
ethical reasoning.

Some writers on ethics have suggested
that it is impossible to distinguish valid and invalid arguments about moral matters C. L. Stevenson maintained this quite recently in his "Ethics and Language" (1944) -- cited by Grice, in "Meaning" (1948).


But it appears that just as inductive argument has been condemned
for failing to conform to the standards of deductive reasoning,
so ethical reasoning is condemned for not conforming to the
standards of either deductive or inductive reasoning.

But
why should it ?

Yet if we do not use such irrelevant standards
of criticism we can surely not condemn ethical reasoning in
general in this way.

For our conception of what is valid
and what is invalid in ethical reasoning must be derived
from a study of ethical reasoning we have here, as in
other spheres, standard examples of valid and invalid argu-
ment. 'Valid' and "invalid " must be used to mark a dis-
tinction within moral reasoning.

It is no more possible for
all ethical arguments to be invalid than for all men to be
small men.

We learn how to use the expression 'valid
argument' with regard to ethics by hearing it applied to
some arguments and not to others.

Our task as philosophers
is to make explicit the principles which we already implicitly
have for distinguishing valid from invalid arguments.

To
query whether any ethical argument is valid is to ask a
pseudo-question, not to raise a serious philosophical pro-
blem.

As S. N. Hampshire said in Mind, 1949, p. 471:

'If the procedure of practical deliberation
does not conform, either in its intermediate
steps or in the form of its conclusions,
with any forms of argument acknowledged as respectable in
logical text-books, this is a deficiency of the logical text-books."


(C) CRITICAL EXAMINATION OF THE FOREGOING
ARGUMENTS

So far we have been concerned to present a reasonably
plausible version of the argument from standard examples
as applied to the question of the general validity of certain
types of argument. I shall now try to show

(a) that the arguments that we have just considered only
appear to be proper examples of the argument from standard
cases because they have been mis-stated.

(b) that when recast in a more correct form they are
indeed the prolegomena to some very important philosophical
investigations, but by no means dispose of the major philo-
sophical problems, as many who have used them have thought
they would. It will, however, be shown that they do require a
change in the traditional formulation of these problems.

(a) The arguments we have been considering are mis-stated

If we ask, absurdly, whether such things as desks are
solid, then the reply can be given that the meaning of 'solid'
is determined by its application to just such things.

As an
explanation of what we mean by 'solid' it would not be
wrong to say 'of the consistency of such things as desks'.


Now if the argument from examples is to be applied to
validity its main contention will be, put in its most succinct
form, that 'valid' means 'like these standard examples in
certain essential respects' and that 'invalid 1 means 'like
these other standard examples in certain essential respects'.


Which arguments are suitable standard cases and what
respects are relevant will of course depend on whether we
are considering deductive, inductive, or ethical arguments.
But this is a mistaken contention, not merely in detail but in
principle, and for a very simple reason.

The reason is that
to call an argument valid is not merely to classify it logically,
as when we say it is a syllogism or modus ponens ; it is at
least in part to evaluate or appraise it ; it is to signify approval
of it.

Similarly to call an argument invalid is to condemn
or reject it.

Therefore while it is plausible to say that
'solid 1 means the same as *of the consistency of certain
standard objects such as desks', it cannot be said that 'valid'
means the same as ' of the same logical character as certain
standard arguments such as a syllogism in Barbara'.

In the
former case we are legitimately equating the meaning of two
classificatory expressions, in the second we are illegitimately
equating the meaning of a classificatory expression, pure and
simple, with an expression which is, at least in part s evaluative
in its meaning.

Grice and Urmson take it that once stated it is obvious that "valid"  is an
evaluative expression.

To speak of a good argument is in most contexts equivalent to speaking of a valid argument, for
example.

It would be ridiculous if, when asked to produce an argument to support a position which I had taken up, were to enquire whether valid or invalid arguments would be preferred.

It seems that any detailed argument on this pont would be otiose.

What we can more importantly do is to consider how the arguments under review can be restated when amended in the light of the point just made and then to ask of them how much they prove in the more correct form.

As a preliminary to the restatement of the argument we must first say something in general about the logic of evaluative terms.

Shortness of space compelled Grice and Urmson to put this portion of their argument very dogmatically, for which fact an apology is undoubtedly due to the reader.

For Urmon's arguments in
support of Urmson's views  the reader is referred elsewhere.
(Urmson,  'On Grading 1 in Mind, 1950 (reprinted in Logic and Language, Vol. II).

Some evaluative expressions claim only to indicate a personal preference, as, for example, when we say that we like something.

One is not compelled to give reasons for liking something, and, if one does, one's reasons are at the worst odd, not improper.

But some evaluative expressions clearly claim a more general validity.

If instead of saying that we like something we say that it is good then reasons are demanded,
are counted as good or bad reasons, and may be argumentatively countered by reasons against.

In general it would seem that the straightforward use of such terms as "good" in
a given field presupposes a set of agreed standards of goodness
in that field amongst those who use it ; giving reasons for or
against a thing being good is to show that it conforms to these
standards.

Thus in a given circle the standards for goodness
in apples may be a certain taste, size, shape, keeping qualities,
absence of worm-holes, etc. ; then to give reasons for saying


"This apple is good."

is to show that it has these characters.

If I sell as good apples which fall far short of the accepted
standards I am liable to get into trouble.

If I do not know
what standards are being used when they speak of good
apples, then in a good sense I do not fully understand them,

There is thus a close logical connexion between an evaluative
expression and the accepted standards for its appropriate use.

But this cannot be identity of meaning, for no evaluation can
be identical in meaning with a description.

Here the analytic-synthetic distinction breaks down, in spite of Grice's and Strawson's attempts to the contrary (vide their "In defense of a dogma", The Philosophical Review, repr. in Grice, "WoW" -- Studies in the Way of Words).

It is surely clear that Valid* is one of those evaluative
terms which, like 'good', claim to show more than a personal
preference. It is more specialized than 'good' in its application, as are * brave' and 'intelligent'.

It is used only to evaluate arguments and then only from a certain point of view
an Invalid argument might indeed be preferable for the persuasion of stupid people, and as a valid argument may have
false premises validity never can involve total satisfactoriness,


But it seems that in its logical character validity resembles
goodness very closely, and when the context is clear we often
use 'valid' and 'good' indifferently.

If there is this resemblance, then we may expect there to be factual criteria or
standards for its use, whether implicit or explicit, and these
criteria will have a close logical connexion not amounting to
identity of meaning with the evaluative term itself.

Let us see
if we can reformulate our arguments from standard examples
in a way which does justice to this logical situation.

If amongst a certain group of people the evaluative distinction between valid and invalid arguments is recognized,
whether the arguments in question be deductive, inductive,
or ethical, then we shall expect to find criteria of validity
which are generally accepted by the group.

Otherwise they
will not be able to use the distinction but only at the best to
argue about how to use it.

If we want to know what these
standards are, we can only find out what these standards or
criteria are if we examine what are agreed to be valid and
invalid arguments and elicit the criteria from them.

Even if
a list of theoretically agreed standards Is already available we
shall still have to check this list against actual practice if we
want to know what the standards actually are ; the argument
from standard examples must teach us that this is the final
court of appeal on the question what the standards are. If it
teaches us nothing else. So, as a reformulation of the argument from standard examples, we can at least say this : if we
can elicit from the usage of a group a set of criteria for the
validity of a certain kind of argument, then it is pointless to
ask whether for that group there is any distinction between
valid and invalid arguments of the kind in question, or to say
that there is no such distinction for them ; for we already
know what the distinction is. But, we can now add, we ourselves are a group which makes such distinctions, so that it is
pointless to ask whether we have, or to deny that we have,
criteria for the validity and invalidity of all these kinds of
argument. Even if we have not yet been able to make these
criteria explicit, none the less the fact that we do succeed in
all these fields in using the Words ' valid' and 'invalid' in an
intelligible way, the fact that there are standard cases of
validity which outside his study no one would deny, shows
that the doubt when expressed in the study is absurd or at
least incorrectly formulated.

What does this argument in fact show ?

Now that we have reformulated the arguments with which we are concerned, or rather indicated a general way of re-
formulating them without elaborating each in turn, what
value can we attach to them ?

Will they convince the philosophical sceptic ?

When we have elicited the standards for counting an apple
good which are current in a group taste, size, absence of
worm-holes, etc., we can no longer ask whether it is possible
for that group to distinguish good and bad apples. But we
can perfectly properly ask why they use these standards and
whether we ourselves have any good reason for using them
a question which in the case of apples is not very difficult to
answer. We may note* too, that the question may be asked in
two quite different spirits. We may ask in a spirit of genuine
doubt whether we should accept these standards, and whether
there are any good reasons for doing so, or we may be quite
happy in the employment of these standards but ask why we
employ them in a spirit of philosophical enquiry.

It is surely clear that in the same way when we have
elicited our standards of validity we shall still be faced with
the further question ; granted that this is the way in which
we distinguish between valid and invalid arguments in this
field, what good reason have we for evaluating arguments in
this way ? Once again s the question can be asked in a spirit
of genuine doubt or as a philosophical enquiry. It is an unfortunate fact that philosophers have continually cast their
question in the form of scepticism, when it is quite clear that
in fact they have no thought of abandoning the distinction.
Now it is a fact of usage that when someone is sceptical of
standards he often formulates his query, not in the form 'Are
there good reasons for using these criteria of goodness ? ', but
in the form 'Are these things really good ?' And so we get
the question ' Are these arguments really valid ? ' One cannot
say that this is incorrect, as those who employ the argument
from standard examples often say, but one can deplore it.

We have the first order question whether, say, an apple is
good, and the second order question why we count such apples
good. It appears that we have the same two questions with
regard to the validity of arguments. We can ask of a par-
ticular argument whether it is good (valid), to which the
answer will sometimes be that it is a paradigm of good arguments.


But we can also ask why we count such arguments
good, and there seems to be nothing improper in asking such
a question, however much we may deplore the pseudo-sceptical
form in which it is often phrased. Professor Ayer has said (in British Empirical Philosophers, pp. 26-7): "in the case of any belief about a matter of fact what
counts as good evidence is inductive evidence. So to raise
the general question whether inductive evidence is good, is to
ask whether what counts as good evidence really Is good evi-
dence ; and I do not think that this question is significant".
This may be correct, so long as we are clear that being inductive Is a criterion of, not what is meant by, good evidence.
But the question still remains why we count inductive evidence
as good evidence ; probably those who ask the question
which Ayer regards as meaningless have In fact been asking
this significant question In a misleading way.

Our arguments show, then, that there is something very
misleading about the formulation of some traditional pro-
blems, and many philosophers have been misled by such
formulations. It is wrong to ask whether Inductive or ethical
arguments are ever valid. But it would appear to be possible
to reformulate these questions to run : * Why do we count as
valid those arguments, inductive or ethical, which we do count
as valid ? ' When so reformulated the questions are quite
proper. It is therefore a mistake to think that the arguments
we have considered dispose of these basic problems, for they
remain.

It is worth noticing that the serious muddles which have
arisen elsewhere have not risen in the case of deductive argument. It is true that some authors have spoken as though
'valid* were here a logical term of the same type as 'disjunctive 1 , but this has not often caused trouble. In the case
of deductive argument the higher order question has never, or
hardly ever, been formulated in the misleading way: 'Are
deductive arguments ever valid', but in the form 'Why are
they counted valid ? J , and in a spirit of philosophical enquiry.
We may not be entirely satisfied with any answer to this
question which has been given, but most discussion of the
topic has been sane, and some of it has surely advanced the
frontiers of knowledge. It is the ridiculous way in which
the higher order question has been put that has prevented
similar sanity with regard to ethics and induction.

The argument from standard examples does not then do
what it was intended to do in the field of validity, at least by
some who have used it. But it does compel us to reformulate
the traditional problems in a healthier way.

This is no matter of pedantry.

In philosophy the correct formulation of problems is half the battle.

To move from the question,

Are any inductive arguments valid?

 to the question,

What good reasons can be given for rating arguments of a certain type
higher than arguments of another type?

is to make a real
advance s before any answer is found.

Above all, we get away
from bogus doubt into methodical philosophical research.
Above all, these arguments compel us to take seriously the
need for a careful analysis of the nature of the inductive,
ethical, and other types of argument that we actually use.
We cannot ask for the reasons for the use of the criteria of
validity that we do use without an actual examination of these
criteria. In the past there has been too much a priori argument about the * possibility of induction * based on equally a
priori notions of what inductive arguments were actually
like.

It is a merit of those philosophers whose arguments we
have been considering that they have seen the need for a careful examination of the forms of our arguments, even if it has
often been a defect that they have not seen that further
enquiries remain.

My main contention has been that the attempt to discuss
the question of validity by means of an argument from
standard examples is misconceived, leading to the attempted
dismissal of genuine philosophical questions. There is indeed
no important philosophical question * Why do we call things
of the consistency of desks solid ? *

If we assimilate the
logical character of validity to that of solidity we are tempted
not to notice, or to dismiss as absurd, the question ' Why do
we count arguments of this and that sort as valid ? ' The
evaluative character of Valid' is here overlooked.

Thus exaggerated claims have been made for the force of these argu-
ments. Such problems as the central 'problem of induction'
have been thought to vanish into thin air. But I have also
wished to claim that these arguments have not been simply
misconceived. There are two problems : what are the criteria
for validity of arguments in a given field, and why do we
employ these criteria. As a result of these arguments we reach
a better formulation of the second problem, and the need for
a thorough examination of the first becomes apparent.

 If it is wrong to deny the existence of the second problem, it is at
least as wrong to fail to notice the autonomy of the first.