This blog will be based on an article done by W.T. Stace. It is called, **Are All Empirical Statements Merely Hypotheses?** It appeared in the philosophical journal known as The Journal of Philosophy Vol. 44, No. 2 (Jan. 16, 1947), pp. 29-38.

It is sometimes stated that all empirical statements are only probable. This was stated by those like, and especially by, Rudolph Carnap. One philosopher who disagreed, and said that some empirical statements are certain, was G.E. Moore. Stace shall agree with Moore, but with some qualifications. The statement that will be the exemplar of what is being talked about will be the statement of “This key is made of iron”. Now this statement is a singular statement like *x *is* Y*.

“To say that this proposition can never be more than probable means, I assume, that there must always be some doubt as to its truth. The question we have to get clear about is: what is the doubt, or what are the doubts, which those philosophers who say that such a statement can never be more than probable, have in mind?”

Some of the doubts could be as follows for what makes this empirical statement probable: the laws of nature are statistical, we could be deceived by some sort of demons or might be dreaming, or statements that we make rely on memory and our memory could be wrong. None of these things seems to be what has lead some to think that all empirical statements are probable. That is because these doubts are arising from *practical doubt* because of the frailty of human faculties.

The philosophers, like Carnap, seem to be relying on *theoretical/logical* doubt. This seems to be based on the logic at which we arrive at empirical truths, regardless of the frailties of particular human beings. They seem to be saying that we arrive at these empirical statements, like “this key is made of iron”, are arrived at by means of induction. And, through the means of induction, we never arrive at certainty by by means of probability.

Stace quotes Carnap on the basic idea of which is to lead to all empirical statements are merely probable. Take the statement that “This key is made of iron”. This proposition will be known as P1. We can test P1 by seeing if it is attracted by a magnet, if it is then we have partial verification of P1. So here is what Rudolph Carnap says, which leads him to state that all empirical statements are merely probable in his book Philosophy and Logical Syntax:

“After that, or instead of that, we may make an examination by electrical tests, or by mechanical, chemical, or optical tests, etc. If in these further investigations all instances turn out to be positive, the certainty of the proposition P1 gradually grows…but absolute certainty we can never attain. the number of instances deducible from P1 is infinite. Therefore there is always the possibility of finding in the future a negative instance.”

Now this is the logical problem that we face. Anytime we perform a new test, and the test is passed, it only adds a degree of probability to the statement that “this key is made of iron”. And the problem, further, is that we can’t completely verify the statement, or be certain of it, because we would have to complete an infinite number of observations. But this is not only practically impossible, it is also logically impossible.

But there is some ambiguity of what Carnap means, because there are two ways that this can be taken. The first thing could be about the different *kinds* of tests. For we noticed that he brought up the tests that could be done, like magnetic, electrical, chemical, and etc. So the it could be meant that the number of different *kind*s of test is infinite, which means we would have to make an infinite number of kinds of tests in order to achieve complete verification of the statements truth. But Stace has an objection to this position.

“If an infinite number of kinds of tests of the key were possible, this would imply that the key must have an infinite number of different characteristics or properties to be tested for. But even if an object can have an infinite number of characteristics, it would not be necessary to test for them all in order to identify the object as iron. All we need is to verify the

definingcharacteristics of iron, which are certainly finite in number. and there is, of course, no logical difficulty about doing that.”

Now there is a second possible meaning for which Carnap has in mind. We could do a single test of a defining characteristic like “being attracted by a magnet”, or what other defining characteristics there might be. These tests only make the statement probable because we may find that the key is attracted one time and perform many of the same tests a thousand times in succession and find the same results as the first test. But we can never be sure that an instance will not turn up in the future in which the object will not be attracted by a magnet (problem of induction). “If the same thing happens in the same circumstances in a vast number of times, each time it happens makes it a little more probable that it will happen again, but it can never be quite certain.”

It is true that scientists perform the same experiments, this is the repeatably of the scientific tests. What one scientist is able to do in a test, it has to be reproducible by other scientists around the world. The same experiment can be repeated by the same experimenter over and over, or can be done by other experimenters around the world. But why are experiments repeated? Is it because each fresh instance of a positive result of the same test adds to the probability of the conclusion? It seems not.

Let us assume that we have an object that is to be tested. We want to test whether it is composed of a certain substance, which we can call X. Now let us suppose that there is only one defining characteristic of X which we call A. The scientist is testing for Y. If Y is found it is a sign that the substance is X. Now, is it true that A may be repeated many times. But why?

“It is not because he supposes that a barren repetition of instances of A makes it more probable that the substance is X. It is always, on the contrary, because he has doubts whether he has satisfactorily established by his observations of the presence of A. It is not the validity of the inductive inference from A to X that he is doubting, but whether A is really present…the doubt which the experimenter is trying to exclude is not any logical doubt about induction, but practical doubts arising from difficulties of observation, possible deficiencies in apparatus, difficulty in ensuring that the experiment is made in the exact conditions required, and so on. He is not doubting that the inductive premises will lead to an absolutely certain conclusion. He is doubting whether he has satisfactorily established the inductive premises.”

What is going on is that the scientist procedure is that a single observation is sufficient to establish an inductive conclusion with certainty. But this is only the case provided that the premises have been established. So it is not the inductive conclusion that is being questioned, but it is the premises that are being questioned. As Stace says, “What is implied by the scientist’s procedure is that a single observation or experiment is sufficient to establish an inductive conclusion with certainty, provided the premises have been established. I hold that the scientist is right.”

Stace locates the problem at three points. And this is the problem of how some philosophers have reached the conclusion that all empirical statements are merely probable.

(1.) One of the problems was how philosophers thought that scientists were repeating experiments to try to dispel logical doubts about the validity of induction. What the scientists were doing, in fact, was trying to dispel practical errors in observing or establishing the premises on which an induction rests. The question of probability doesn’t fall within the inductive argument, but outside of the inductive argument.

“That is to say, what is only probable is not that,

ifA is once associated with B, it will always be associated with B, but that A has actually been found associated with B; not thatifa substance has a certain specific gravity it is gold, but that the substance now before me actually has that specific gravity…a natural mistake located the question of probability within the inductive argument instead of outside of it; have extrapolated it from the practical sphere of observation, measurement, and so on, where it actually belongs, to the logical sphere of the inductive inference in which in reality it has no place.”

So the problem is not in the inductive argument itself, but outside of the argument. What is outside of the argument is making sure that you have made an observation that meets with the premises of the argument. This is what constant testing is about, to make sure that the observations are in line with the premises. It is not the argument being questioned, but something outside of the argument that is being questioned.

(2.) Another reason that it seems that it is brought up that empirical statements are probable deals with the view of induction where an application of the inductive principle to a type of cases different from that of the Iron key. This other application is based on generalizing from observations. For example, we generalize from observations of a number from a certain class to the whole class. This means, from observing some white swans, we go on to generalize to the class of swans. From seeing a certain number of swans being white, and not observing any black swans, we go on to say that All swans are white. This will be dealt with a little later on.

(3.) This view seems to follow, as some philosophers think, from what David Hume had to say on the problem of Induction. Hume showed that we can’t “prove” a conclusion in an inductive argument. Because of this, some seem to have imagine that because we can’t prove it, we can at least make it probable. But it doesn’t seem that this follows from what Hume said on the problem of Induction. But Stace does think that something follows from what Hume said on this problem.

Imagine that we have a single instance of A being associated with B, and we’ve ruled out all practical doubts from possible errors of observation or experiment. We now have, logically, two positions that we can take up.

The first is that we can assume the validity of the principle of Induction. So, in this single instance, we can conclude that A is always associated with B, and our conclusion follows with absolute certainer from our two premises of single observed association of A with B *and* the principle of induction. With these two premises, the conclusion is certain to start with, and so there is no increasing probability or probability at all.

The second is that you may not assume that validity of the inductive principle. Now this means that we follow Hume, which means that there’s no logical connection between the premises and the conclusion of induction. This means, nothing follows from induction, neither certainty nor probability. No matter how many single instances that support our inductive conclusion, the probability never arises above zero. (Karl Popper would agree with this point). There is no connection to say that because the conclusion obtained, that we can say that the probability of the premises rises some more. They are disconnected. It is like having three dots on a sheet of paper. They are disconnected from each other. So when we affirm one, we can’t affirm any of the others because they’re not connected with one another.

“I have affirmed that, given the inductive principle, a single case will prove the inductive conclusion with certainty, I ought to give a formulation to the inductive principle which embodies this…”

If in even a single instance, we have observed that a thing of the sort A is associated with a thing of the sort B, then on any other appearance of A,.””provided the other factors present along with A are the same on both occasions, it is certain that A will be associated with B

There is the clause of “provided the other factors present along with A are the same on both occasions.” This forms part of the principle, which comes down to “Same cause, same effect”. There is an example to help make this point clear. If the bell is struck in air then it produces sound. But it doesn’t follow that a bell struck in a vacuum will produce sound. This is because of the clause that was inserted into the principle. The factors aren’t the same, and so they’re not the same type of thing. But it does introduce a new inductive discovery.

There is one obvious objection that one could make to this principle. It could be said that this new interpretation is merely an assumption that is incapable of proof. So if this is a matter of being arbitrary choice of how to formulate it in terms of certainty and probability, then we ought not to assume more than is necessary to justify our sciences and our practice. So someone could say, “it will be quite sufficient for these purposes to assume that, if A is associated with B now, it will probably be associated with B at other times and places. On this ground the probability formulation should be preferred.”

But putting the term certainty in there is not meant to be arbitrary, but it is mean to represent a formulation of the assumption which has been the basis of science and practice. But maybe Stace should be more clear, which is what he tries to do like as follows:

“If you have one case of a set of circumstances A associated with B, and you are quite sure you have correctly established this one association, then, assuming the uniformity of nature, or the reign of law, or the principle of induction-call it what you will- a repetition of identically the same set of circumstances A is bound to be associated with B. For if not, you would have a capricious world, a world in which A sometimes produces B, and sometimes it does not, a world in which the kettle put on the fire may boil today, but freeze tomorrow. And this would clearly be a violation of the principle of induction which you have assumed.”

Now, if you assume the principle of induction, then a single case validates an induction. But now Stace will try to prove his second contention that if you don’t assume the principle of induction, your inductive conclusion aren’t probable at all and there’s no repetition of instances, so no matter how great the number, then the probability is never raised above zero.

To establish this position, Stace will assume that Hume is right. This means, between the premises and the conclusion of an inductive argument there is absolutely no logical connection at all. This means that there is nothing to establish the slightest probability because they’re is no connection between them. So if we affirm one part, it has no connection to another to raise the probability of this part that is connected to what we affirmed. They are so completely disconnected that there’s no logical connection to even bring up probability.

For example, here is what Al-Ghazali said about causality, which is the same position that David Hume took up, and this is based in some ways on the principle of induction. “The affirmation of one does not imply the affirmation of the other; nor does its denial imply the denial of the other. The existence of one is not necessitated by the existence of the other; nor its non-existence by the non-existence of the other.” So when we affirm one thing with induction, like a correct experiment, this in no way can increase any probability when the affirmation of one doesn’t imply the affirmation of the other. How can you raise the probability when what you affirm has no connection to anything else to raise the probability of this other thing? You can’t.

Stace goes on to try to examine the types of cases in which generalize a whole class from a number of instances that are smaller than the whole class. Try to generalize about a whole class of swans from observing a few of the swans that are suppose to make up the whole class. If we observe one swan and it is white,nto conclude that all swans are white, we might be accused of generalizing from one instance. But if we make 10,000 observations, we might think we have a degree of probability to support the generalization. We go on to make observe 1 billion swans and they were white. This might lead us to go on to admit that the hypothesis has become even more probable. So, someone might say to defend the probability view, that how can we deny that we probability and use the probability view of induction?

“But the inductive principle only holds with the proviso, “if the factors present along with A are the same” in subsequent repitition of A. And this case of the swans is simply a case in which it is extremely difficult to be sure that this is so. A in this case means the defining characteristics of the class swan, and B means whiteness. Now different swans will have, along with the defining characteristics A, a number of other characteristics. and these will differ with different individual swans, not to mention circumambient differences of environment. Thus the first case of A you observed was really ACDE, and this was associated with B. The second case was APQR, the third AXYZ. Now, of course, it does not follow from the principle of induction that because ACDE was associated with B, therefore APQR and AXYZ must be associated with B. For we do not have there that exact repetition of the same sets of circumstances which the inductive principle requires.”

To try to remedy the situation that we are in, we constantly repeat observations of this class of swans. Now if we keep making these observations of A, and they’re found to have B, then we think it becomes more and more likely that we have eliminated other certain possibilities, and raise the probability. We want to eliminate some of the accidental characteristics of certain swans. This would be something like they’re size. food they eat, and the climates that they live in. When we rule out sets of circumstances as irrelevant, they become more probable.

The fundamental reason why there is constant repetition of observation on new members of class is that although in theory the association of A with B, once it is observed must always hold, is because in practice we never get our cases of pure A. “We can not isolate the system. It is always mixed up with extraneous circumstances. Thus the doubt which we are trying to dispel by repeated observations has nothing at all to do with Hume’s doubt about the validity of induction…” That doubt can’t be dispelled, no matter now many numerous observations we make. But the doubt that we are trying to get rid of isn’t the logical doubt. The doubt we are trying to get rid of is the practical doubt from the enormous complexity of nature, our frailty of our intellects which are unequal with the task to disentangle the complexities, or the inadequacy of the instruments that we have at our disposal to isolate the system present.

Some, like Carnap, have divided knowledge into empirical knowledge and necessary propositions. Necessary propositions would be those like mathematics and logic. Now the empirical propositions could be considered doubtful because the practical doubts that arise from our human infirmities. But this means that we ought to have the same doubts in concern with mathematics. This is why we have people that check our work in mathematics, to make sure that we made no practical doubts in the process that we followed.

“There is one sense in which mathematical, or, in general, deductive conclusions are certain this may be called the logical or theoretical sense. And there is another sense, which may be called the practical sense, in which they are only probable, since the mathematician or the syllogizer may err in his reasoning. The mathematician may miscalculate, and the syllogizer may make any one of a hundred mistakes. And if practical doubts are not a ground for denying that, in an appropriate sense, mathematics is certain, then practical doubts can not be a ground for denying that, in an appropriate sense, empirical conclusions are uncertain.”

“As it is with mathematical truths, so precisely it is with empirical truths. There is one sense in which an inductive conclusion is certain, namely, the theoretical sense that it follows with certainity from a single observation plus the inductive principle. And there is another sense, the practical one, in which it is probable only, because there may be errors in observation, experimentation, and the like.”

“The statement that empiricial knowledge may be theoretically certain is, of course, subject to the proviso that we accept the inductive principle. If we don’t accept it, then, of course, empirical knowledge is not even probable. It has no validity at all. In no case does any question of probability enter into the matter.”