How are Hypotheses Tested?
According to Wesley Salmon, test their hypotheses by deducing observational consequences from them. This suggests a degree of independence of the hypothesis from observations or from testing/experiments, etc. Indeed it suggests, quite simply, that the hypothesis comes before all observations or tests. An isn't that how the layperson, after all, takes the word hypothesis? He would say that the whole point of a hypothesis is that it's a stab in the dark.
Yet if a hypothesis, scientific or otherwise, were really a complete stab in the dark, what would be the point of it? It would be arbitrary as well as a complete fabrication. It would be plucked from the air and would therefore, surely, have no relevance to anything scientific. For example, I can formulate the hypothesis that the sun is made of cheese. What's the point of it? It's clearly false and clearly scientifically illiterate. Nonetheless, the suggestion seems to be that this is what a hypotheses is – a complete stab in the dark.
Thus we can conclude one thing: even though an hypothesis comes before its observational consequences, it clearly doesn't come before all previous observations. It clearly doesn't come before a whole lot more as well: previous theories, laws, tests, experiences, etc. To use the language of old-style epistemology: an hypotheses is not aprioristic. If it were truly a priori, in the traditional rationalist sense of that term, then it wouldn't be a scientific hypothesis (though I suppose it could be a non-scientific hypothesis).
Anyway, the hypothesis is tested in the sense that the observations it predicts, or the observational consequences of its content, do come to pass.
Now we meet two more scientific terms: confirmation and disconfirmation.
If the predicted observations or events occur, then hypothesis is confirmed. If they don't, it is disconfirmed.
I mentioned that no hypothesis can be truly free-standing. Another reason for this is that the formulation of the hypothesis must assume various or many things. These assumed things have been called auxiliary hypotheses because they underpin the given hypothesis. Or, alternatively, the auxiliary hypotheses are contained within the (new) hypothesis.
Wesley Salmon gives the example of a medical experimenter who predicts that a certain bacillus will be found in the blood of a certain organism. Now in order to be scientifically sure or certain of his experiment or hypothesis, he must accept certain auxiliary hypotheses about the optics which are part of the experimental set-up and which includes the microscope itself. However, in actual fact, the scientist in practice doesn't really need to assume or even accept any hypotheses about optics at this - or any - moment in time. He can ignore them. He's not a scientist of optics nor an expert on microscopes. Other scientists are. What this really means is that both logically and scientifically these auxiliary hypotheses underpin his experiment even if the medical experimenter need not - and probably will not – know a thing about these extra hypotheses: he hasn't got the time.
This is an example of Quine's scientific holism and the scientist concerned need not be aware of the “web of science” or even large parts of it. The web exists regardless of the particular knowledge of the individual scientist. It existed before him, during the experiment and it will exist after the experiment.
It may seem strange to argue that “a true observational consequence follows deductively from a given hypothesis”. Or, more precisely, it is the use of the word “deductively” that may seem strange. Surely deduction is a purely logical matter? How can anything “observational” deductively follow from, well, anything logical or even from anything non-logical? Surely only theorems, conclusions, etc. can deductively follow from something or even from a given hypothesis. That is an understandable position.
Nonetheless, if the hypothesis has a certain given content, and that content says that if H then O, then if H then O it doesn't matter if the content of O includes predictions about observations, experience or that which is empirical. After all, this is in fact a conditional statement. That is, if this hypothesis is correct, then O will occur. It doesn't matter if the hypothesis, or O, has empirical content, or says that something observational will occur. It says that if it were correct, then there would be certain observational consequences. The hypothesis, or conditional, generates, as it were, what deductively follows from it, even if what deductively follows from it are indeed observational consequences. In other words, the hypothesis is not claiming that there are logically deductive sequences in nature, as it were. It's saying that given hypothesis H, then O (the observations consequences) will follow. The deductive relationship is between H and O, not between one aspect of the world and another (in a non-Humean manner, as it were). Alternatively, there is no necessity in the world, but there is a certain kind of necessity, or at least a deductive consequence, from H to O.
Hypothesis and Evidence
An argument about the independence of an hypothesis from its evidence (or from evidence generally), or from observational consequences, can be articulated by saying that given exactly the same evidence, various and many hypotheses can explain that same evidence. Basically, this is a way of making the obvious point that evidence and hypothesis are not the same thing. Alternatively, an hypothesis is more than the evidence which supports it. (This is basically a rephrasing of the idea that “theory is always underdetermined by all available and relevant evidence”.)
In fact, just as I stated the truism that hypothesis and all available and relevant evidence are not the same thing, and also that rival hypotheses can explain the same evidence, so too it is the case that all these rival hypothesis - which are fighting to explain the same evidence - are not equal or identical either. And that lack of identity or equality, again, has nothing to do with the relevant evidence (which is the same for all the rival hypotheses).
It is commonly said, by both scientists and philosophers of science (though less by the latter), that these other factors include the degree of simplicity of the hypothesis as well as its explanatory power, esthetic value, comprehensiveness, etc. However, even though I have stressed the fact that evidence is not everything, it is, obviously, of vital importance. (How could it not be in science?) So it is true that many commentators, not always scientists, have stressed that Watson and Crick were esthetically delighted by the beauty of the double helix hypothesis for the structure of the DNA molecule. That's true. However, if they had wanted purely artistic pleasure they would have become painters or composers. That esthetic pleasure was largely generated by the simplicity of the double helix hypothesis. But simplicity, in scientific theory, is not an end in itself. That simplicity, in this and in many other cases, meant that there was a good chance that the said hypothesis is true/correct. That is, simplicity generated beauty and that beauty/simplicity generated the strong possibility of a correct hypothesis.