Monday, September 3, 2012

Constructive empiricism vs realism in philosophy of physics

To: Christopher F.
From: Geoffrey Klempner
Subject: Constructive empiricism vs realism in philosophy of physics
Date: 4th April 2008 11:10

Dear Chris,

Thank you for your email of 24 March, with your essay towards the Associate Award entitled, 'Constructive Empiricism vs Realism in the Philosophy of Physics'.

In answer to your proposal, I would be very interested in a new Pathway. As it happens, I am currently writing one myself, a ten unit course 'Ethical Dilemmas'. See and

The Pathways programs are a bit more than 'overviews'. The kind of thing I would be looking for is a cross between an introductory text and a monograph: in other words, this would be an opportunity for you to develop a general view of problems within this area while at the same time leading the student through the topics that they need to grasp in order to appreciate the view that you are presenting.

Hopefully, this would, if anything, make the project more rather than less attractive to you!


This is a very competent and knowledgeable piece of work. However, there are some issues I want to take up in particular with your interpretation and application of the Quine-Duhem thesis.

The point that a scientific theory has 'auxiliary hypotheses' is not necessarily an application of the Quine-Duhem thesis. It is a criticism which has been levelled specifically at Popper's account of the demarcation between science and non-science in terms of falsifiability. You make the point that when Newton saw that there were slight deviations from the paths of the planets predicted by his mechanics he didn't give up the theory; rather, he assumed (rightly, as it turned out) that there were unobserved entities exerting some influence on planetary orbits.

The general point is one which overlaps with Kuhn's account of 'normal science': it is good methodology to take a conservative view when faced with apparently contrary experimental results. If a theory has proved fruitful in the past, you don't throw it overboard but do your best to save it, and only give it up as a last resort.

This is indeed fully consistent with the *rejection* of the thesis of the underdetermination of theory by data. We are investigating in a state of ignorance where we are nowhere close to having 'all the data' (whatever that would mean). For all we know, if we did have all the data there would be no underdetermination, but on the contrary only one theory which fitted. But it would still be good methodology to attempt to save theories which work well rather than dumping them.

The second point concerns your example of the different formalisms for mechanics. I don't have sufficiently good knowledge of mechanics to know exactly how the different formalisms work, but my strong hunch (correct me if I'm wrong) is that the situation is similar to that in formal logic, where we have a choice between equivalent systems with different sets of axioms, and indeed systems which do not have axioms but rather rely on rules of deduction.

As applied to mechanics, this is a strong argument against 'entity realism' at least to positing entities which correspond, in a realist sense, to the primary concept of a particular formalism. That's not how they work. However, just as one would say that 'first-order predicate calculus' is essentially the same abstract structure, irrespective of whether you are using an axiomatic or a natural deduction system, so the different formalisms of mechanics are all formalisms of the very same science of *mechanics*, and their differences merely matters of practical utility for solving given problems.

This still isn't the Quine-Duhem thesis. Once again, it would be fully consistent with the rejection of the Quine-Duhem thesis to allow the possibility of alternative formalisms for mechanics.

What, then, is the Quine-Duhem thesis? The thesis is a generalization of the anti-falsifiability point, which states that we can *always* save a theory, no matter what. No empirical discovery ever forces us to give up a statement provided that we are prepared to make sufficiently large adjustments elsewhere. However, as Quine would argue, there are pragmatic constraints which in a real sense limit the number of fruitful ways of responding to given evidence. It is just plain nutty to attempt to defend the flat earth theory against all contrary observations, although in principle (if Quine is right) it can be done. This is pragmatism, with teeth.

How does this impact on realism? You might think, pretty obviously, you can't be a militant pragmatist and a realist. But I'm not so sure. Traditional pragmatism was ranged against absolute idealism (e.g. the famous debate between William James and F.H. Bradley). There, it was pretty clear that what was at stake was a metaphysics, a view about the ultimate nature of reality. Here, by contrast, we are talking about methodology. Can't you be a realist with a pragmatist methodology? You might think that the realism is otiose, but then again we have the example of Popper who is convinced of the existence of his 'third world' (the objective truth) while making no real practical use of this belief in describing his methodology of falsificationism.

All the best,