Abduction

Reasoning to whatever would best account for what you have observed - the only mode that generates new ideas.

Theme: Knowledge, truth and reason
Tradition: Pragmatist
Term fixed: 1878

In plain terms

Deduction works out what follows. Induction generalises from cases. Neither of them ever produces a new idea - they only rearrange or extend what is already in play.

Abduction is the third thing. You have a surprising fact, and you ask what would have to be true for it to be unsurprising. The answer is a hypothesis, and it arrives from nowhere in particular.

It is the mode used by every diagnosis, every detective, every mechanic and every scientist proposing anything. It is also the weakest of the three, because there is always more than one explanation and best is doing a great deal of unexamined work. Simplest, most familiar and most flattering are all candidates for what it means in practice.

As a philosopher would put it

Peirce places abduction as the first stage of enquiry: it generates hypotheses, deduction draws out their consequences, and induction tests them. His formulation is that a surprising fact is observed, and if some hypothesis were true the fact would be a matter of course, so there is reason to suspect the hypothesis.

Harman's inference to the best explanation is broader and does the epistemological work. It licenses believing the hypothesis that best explains the evidence, and is presented as underlying most ordinary knowledge - other minds, the past, the external world.

The standard objection is van Fraassen's bad lot argument: we can only choose the best of the explanations we have thought of, and there is no reason to think the true one is among them. The second objection concerns the criteria. Simplicity, scope and fit with background belief are the usual candidates, and it is not clear why any of them should track truth rather than convenience - which is the anti-realist's opening in the debate over scientific realism.

Where you meet it

In 1982 Barry Marshall and Robin Warren, in Perth, were looking at gastric biopsies. Warren had found curved bacteria in the stomach lining of patients with gastritis, in a place where the medical consensus held that nothing could live because of the acid.

The standard explanation of ulcers was stress and excess acid. Marshall and Warren proposed a different one: the bacteria were causing the inflammation, and the bacteria were there because they had adapted to the environment rather than because they had wandered in. It explained the biopsies, the pattern of recurrence after acid-suppressing treatment, and the clustering of cases.

The reception was poor and their papers were rejected. In 1984 Marshall drank a culture, developed gastritis, and had the organism recovered from his own stomach. That is not what made the hypothesis good - the explanation was already the better one - but it is what made it heard. They received the Nobel Prize in 2005.

What is actually disputed

Whether the best explanation is likely to be true.

Realists take explanatory success as evidence, and argue that the success of science would otherwise be a miracle. Van Fraassen's reply is that we compare only the hypotheses that have occurred to us, and that being the best of a bad lot is no recommendation. The realist rejoinder - that the same worry applies to every inference and does not stop us reasoning - is a standoff rather than a resolution.

A second dispute is whether abduction is a distinct mode at all. Bayesians argue it dissolves into probabilistic updating, with explanatory virtues functioning as priors. Peirce insisted it is irreducible because it is the only mode that introduces a hypothesis rather than evaluating one, and the Bayesian machinery has nothing to say about where hypotheses come from - which is a real gap and possibly not a philosophical one.

Commonly confused with

deduction in detective fiction What Holmes calls deduction is almost always abduction. He proposes the hypothesis that would account for the traces.
Induction c. 330 BC Induction generalises from instances of the same kind. Abduction posits something of a different kind entirely, which is what makes it creative.
confirmation bias Preferring the explanation you already favour is abduction done badly. The mode itself requires comparing candidates.

Defined against

Deduction Deduction extracts what is already present; abduction introduces something that was not, which is why only one of them can be wrong in an interesting way

Philosophers

Charles Sanders Peirce Names it, and argues it is irreducible to the other two
John Dewey Enquiry as beginning from an indeterminate situation and proposing a resolution
Karl Raimund Popper Denies that hypothesis generation has any logic to it
William James The pragmatist context in which hypothesis and test are one process

Works

How to Make Our Ideas Clear Where the three modes are laid out
Experience and Nature Enquiry as hypothesis in a situation
The Logic of Scientific Discovery The denial that discovery has a logic
The Structure of Scientific Revolutions Where explanation choice is shown to depend on the framework
Pragmatism Hypothesis and consequence as one movement

Schools

Classical Pragmatism Where the mode is identified and defended
Vienna Circle Where the context of discovery is excluded from logic
Cambridge Analysis Where the formal treatment of confirmation is attempted

Related concepts

Inference c. 330 BC · Induction c. 330 BC · Explanation 1948 · Knowledge c. 369 BC · Intuition 1641

Peirce named it and argued it was a third mode irreducible to the other two; the phrase inference to the best explanation is Gilbert Harman's, from 1965, and the two are not quite the same notion.