When I was a grad student in psychology, I audited a 20th century philosophy of science course. I did my best to absorb all the great discussions, but one remark the professor made continues to bother me because I've never been able to make sense of it: She said that a belief that science is continually (if asymptotically) improving does not necessarily mean that there is a true state of affairs that is being approximated. Yet, my understanding of fallibilism is that in order for a theory to be more or less accurate, there must be some criterion against which the theory is compared to determine its accuracy -- thus, fallibilism appears to carry an ontological commitment that some state of affairs actually is the case, even if we will never apprehend it perfectly.

The only way I've been able to make any sense of my professor's claim is to understand it as rejecting the premise that the world can ever be described as a "state of affairs" -- in other words, that the world our theories seek to describe is in constant flux (hat tip to Heraclitus). But I don't find this explanation very satisfying.

2 Answers 2


Your interest is in fallibilism and its connection to accuracy. The quotation from the professor, however, is about improvement and its relationship to approximate truth.

If you understand improvement as necessarily involving increasing accuracy, then yes, improving scientific representations are those that improve in accuracy. And yes, increasing accuracy presupposes the possibility of inaccuracy, which is to say error or fallibility.

However, note that there are other ways of understanding improvement, which, the professor indicates, do not necessarily involve accuracy to a state of affairs. If science serves as an increasingly good tool for solving problems and interacting with the world, it's improving, and we can observe that without ever discussing accuracy to states of the world. Scientific realists will offer that such accuracy is itself the best explanation of science working better, but that's a further inference. Besides problem-solving, we might also understand improvement in science as a function of increasingly successful prediction and retrodiction (or backwards prediction), or along other dimensions like simplicity, generality, complexity, applicability, and understandability.

To emphasize the answer to your highlighted question, though, yes, I think fallibilism, or allowance of the possibility of error, requires that we accept that there might be something about which we can be in error, and thus involves at least this minimal ontological commitment. We could go on to discuss how minimal it can be!

  • Thank you for the reply; it does clarify some of my confusion. I understand that we can treat scientific progress as itself an empirical observation in need of an explanation -- and that scientific realism is one very intuitive such explanation. As you say, however, this is a further inference -- but I'd like to push on this point a bit. Identifying as a Jamesian pragmatist, I would say that I endorse the scientific realist explanation because it works. But are there any alternative, even minimally compelling explanations for the progress of science (other than the Cartesian demon)? Commented Mar 27, 2015 at 18:04

Fallibilism can be applied to two concepts of science - it's predictions and it's representations.

It is a prediction that if I let go of the book in my hand that it will fall towards the floor; this is truth in lower-case; and unproblematic in terms of fallibilsm.

It is a representation that this is by the 'force of gravity' or by the 'curvature of spacetime'' or by 'entropic gravity' - this is truth in Capital Letters; and here fallibilsm is more problematic; by what consideration can we say that this representation is reality as it is?

An illustration may make this clear: a video shot of the book falling to the floor is a faithful representation of this event; but no-one would mistake this film for reality.

So to be precise, fallibilsm operates on the order of a representational rather than ontological commitment.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .